{ "best_metric": 67.7750577995376, "best_model_checkpoint": "./whisper-small-th-ver2/checkpoint-25000", "epoch": 7.759456838021339, "global_step": 25000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.0, "loss": 1.5771, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.0000000000000003e-06, "loss": 1.6984, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.000000000000001e-06, "loss": 1.2741, "step": 3 }, { "epoch": 0.0, "learning_rate": 4.000000000000001e-06, "loss": 1.6695, "step": 4 }, { "epoch": 0.0, "learning_rate": 6e-06, "loss": 1.346, "step": 5 }, { "epoch": 0.0, "learning_rate": 8.000000000000001e-06, "loss": 1.4355, "step": 6 }, { "epoch": 0.0, "learning_rate": 1e-05, "loss": 1.3573, "step": 7 }, { "epoch": 0.0, "learning_rate": 9.999599919983998e-06, "loss": 0.8819, "step": 8 }, { "epoch": 0.0, "learning_rate": 9.999199839967994e-06, "loss": 1.3703, "step": 9 }, { "epoch": 0.0, "learning_rate": 9.998799759951991e-06, "loss": 0.9687, "step": 10 }, { "epoch": 0.0, "learning_rate": 9.998399679935988e-06, "loss": 0.8653, "step": 11 }, { "epoch": 0.0, "learning_rate": 9.997999599919986e-06, "loss": 0.7996, "step": 12 }, { "epoch": 0.0, "learning_rate": 9.997599519903983e-06, "loss": 0.731, "step": 13 }, { "epoch": 0.0, "learning_rate": 9.997199439887978e-06, "loss": 0.5532, "step": 14 }, { "epoch": 0.0, "learning_rate": 9.996799359871974e-06, "loss": 0.6399, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.996399279855971e-06, "loss": 0.8593, "step": 16 }, { "epoch": 0.01, "learning_rate": 9.99599919983997e-06, "loss": 0.5994, "step": 17 }, { "epoch": 0.01, "learning_rate": 9.995599119823966e-06, "loss": 0.5885, "step": 18 }, { "epoch": 0.01, "learning_rate": 9.995199039807963e-06, "loss": 0.7489, "step": 19 }, { "epoch": 0.01, "learning_rate": 9.99479895979196e-06, "loss": 0.479, "step": 20 }, { "epoch": 0.01, "learning_rate": 9.994398879775956e-06, "loss": 0.5809, "step": 21 }, { "epoch": 0.01, "learning_rate": 9.993998799759953e-06, "loss": 0.5514, "step": 22 }, { "epoch": 0.01, "learning_rate": 9.99359871974395e-06, "loss": 0.6545, "step": 23 }, { "epoch": 0.01, "learning_rate": 9.993198639727946e-06, "loss": 0.6559, "step": 24 }, { "epoch": 0.01, "learning_rate": 9.992798559711943e-06, "loss": 0.6718, "step": 25 }, { "epoch": 0.01, "learning_rate": 9.99239847969594e-06, "loss": 0.8164, "step": 26 }, { "epoch": 0.01, "learning_rate": 9.991998399679936e-06, "loss": 0.5105, "step": 27 }, { "epoch": 0.01, "learning_rate": 9.991598319663933e-06, "loss": 0.5905, "step": 28 }, { "epoch": 0.01, "learning_rate": 9.99119823964793e-06, "loss": 0.6225, "step": 29 }, { "epoch": 0.01, "learning_rate": 9.990798159631926e-06, "loss": 0.4422, "step": 30 }, { "epoch": 0.01, "learning_rate": 9.990398079615925e-06, "loss": 0.6029, "step": 31 }, { "epoch": 0.01, "learning_rate": 9.989997999599921e-06, "loss": 0.5947, "step": 32 }, { "epoch": 0.01, "learning_rate": 9.989597919583918e-06, "loss": 0.6291, "step": 33 }, { "epoch": 0.01, "learning_rate": 9.989197839567915e-06, "loss": 0.6588, "step": 34 }, { "epoch": 0.01, "learning_rate": 9.988797759551911e-06, "loss": 0.5866, "step": 35 }, { "epoch": 0.01, "learning_rate": 9.988397679535908e-06, "loss": 0.5337, "step": 36 }, { "epoch": 0.01, "learning_rate": 9.987997599519905e-06, "loss": 0.5947, "step": 37 }, { "epoch": 0.01, "learning_rate": 9.987597519503901e-06, "loss": 0.4872, "step": 38 }, { "epoch": 0.01, "learning_rate": 9.987197439487898e-06, "loss": 0.6981, "step": 39 }, { "epoch": 0.01, "learning_rate": 9.986797359471895e-06, "loss": 0.6554, "step": 40 }, { "epoch": 0.01, "learning_rate": 9.986397279455892e-06, "loss": 0.6752, "step": 41 }, { "epoch": 0.01, "learning_rate": 9.985997199439888e-06, "loss": 0.4698, "step": 42 }, { "epoch": 0.01, "learning_rate": 9.985597119423885e-06, "loss": 0.5523, "step": 43 }, { "epoch": 0.01, "learning_rate": 9.985197039407882e-06, "loss": 0.7826, "step": 44 }, { "epoch": 0.01, "learning_rate": 9.98479695939188e-06, "loss": 0.5894, "step": 45 }, { "epoch": 0.01, "learning_rate": 9.984396879375877e-06, "loss": 0.4831, "step": 46 }, { "epoch": 0.01, "learning_rate": 9.983996799359873e-06, "loss": 0.4343, "step": 47 }, { "epoch": 0.01, "learning_rate": 9.98359671934387e-06, "loss": 0.5412, "step": 48 }, { "epoch": 0.02, "learning_rate": 9.983196639327867e-06, "loss": 0.6191, "step": 49 }, { "epoch": 0.02, "learning_rate": 9.982796559311863e-06, "loss": 0.7007, "step": 50 }, { "epoch": 0.02, "learning_rate": 9.98239647929586e-06, "loss": 0.5049, "step": 51 }, { "epoch": 0.02, "learning_rate": 9.981996399279857e-06, "loss": 0.5184, "step": 52 }, { "epoch": 0.02, "learning_rate": 9.981596319263853e-06, "loss": 0.6384, "step": 53 }, { "epoch": 0.02, "learning_rate": 9.98119623924785e-06, "loss": 0.3895, "step": 54 }, { "epoch": 0.02, "learning_rate": 9.980796159231847e-06, "loss": 0.575, "step": 55 }, { "epoch": 0.02, "learning_rate": 9.980396079215844e-06, "loss": 0.6402, "step": 56 }, { "epoch": 0.02, "learning_rate": 9.97999599919984e-06, "loss": 0.4839, "step": 57 }, { "epoch": 0.02, "learning_rate": 9.979595919183837e-06, "loss": 0.7014, "step": 58 }, { "epoch": 0.02, "learning_rate": 9.979195839167835e-06, "loss": 0.4743, "step": 59 }, { "epoch": 0.02, "learning_rate": 9.978795759151832e-06, "loss": 0.4956, "step": 60 }, { "epoch": 0.02, "learning_rate": 9.978395679135829e-06, "loss": 0.7403, "step": 61 }, { "epoch": 0.02, "learning_rate": 9.977995599119825e-06, "loss": 0.64, "step": 62 }, { "epoch": 0.02, "learning_rate": 9.977595519103822e-06, "loss": 0.49, "step": 63 }, { "epoch": 0.02, "learning_rate": 9.977195439087817e-06, "loss": 0.493, "step": 64 }, { "epoch": 0.02, "learning_rate": 9.976795359071815e-06, "loss": 0.3851, "step": 65 }, { "epoch": 0.02, "learning_rate": 9.976395279055812e-06, "loss": 0.5379, "step": 66 }, { "epoch": 0.02, "learning_rate": 9.975995199039809e-06, "loss": 0.6384, "step": 67 }, { "epoch": 0.02, "learning_rate": 9.975595119023805e-06, "loss": 0.6116, "step": 68 }, { "epoch": 0.02, "learning_rate": 9.975195039007802e-06, "loss": 0.6796, "step": 69 }, { "epoch": 0.02, "learning_rate": 9.974794958991799e-06, "loss": 0.5258, "step": 70 }, { "epoch": 0.02, "learning_rate": 9.974394878975796e-06, "loss": 0.8085, "step": 71 }, { "epoch": 0.02, "learning_rate": 9.973994798959792e-06, "loss": 0.4964, "step": 72 }, { "epoch": 0.02, "learning_rate": 9.97359471894379e-06, "loss": 0.479, "step": 73 }, { "epoch": 0.02, "learning_rate": 9.973194638927787e-06, "loss": 0.5513, "step": 74 }, { "epoch": 0.02, "learning_rate": 9.972794558911784e-06, "loss": 0.4583, "step": 75 }, { "epoch": 0.02, "learning_rate": 9.97239447889578e-06, "loss": 0.5491, "step": 76 }, { "epoch": 0.02, "learning_rate": 9.971994398879776e-06, "loss": 0.6476, "step": 77 }, { "epoch": 0.02, "learning_rate": 9.971594318863772e-06, "loss": 0.6011, "step": 78 }, { "epoch": 0.02, "learning_rate": 9.97119423884777e-06, "loss": 0.4841, "step": 79 }, { "epoch": 0.02, "learning_rate": 9.970794158831767e-06, "loss": 0.5963, "step": 80 }, { "epoch": 0.03, "learning_rate": 9.970394078815764e-06, "loss": 0.6128, "step": 81 }, { "epoch": 0.03, "learning_rate": 9.96999399879976e-06, "loss": 0.5768, "step": 82 }, { "epoch": 0.03, "learning_rate": 9.969593918783757e-06, "loss": 0.6879, "step": 83 }, { "epoch": 0.03, "learning_rate": 9.969193838767754e-06, "loss": 0.5549, "step": 84 }, { "epoch": 0.03, "learning_rate": 9.96879375875175e-06, "loss": 0.3639, "step": 85 }, { "epoch": 0.03, "learning_rate": 9.968393678735748e-06, "loss": 0.5598, "step": 86 }, { "epoch": 0.03, "learning_rate": 9.967993598719746e-06, "loss": 0.4022, "step": 87 }, { "epoch": 0.03, "learning_rate": 9.967593518703743e-06, "loss": 0.4302, "step": 88 }, { "epoch": 0.03, "learning_rate": 9.967193438687738e-06, "loss": 0.7108, "step": 89 }, { "epoch": 0.03, "learning_rate": 9.966793358671734e-06, "loss": 0.6595, "step": 90 }, { "epoch": 0.03, "learning_rate": 9.966393278655731e-06, "loss": 0.5669, "step": 91 }, { "epoch": 0.03, "learning_rate": 9.96599319863973e-06, "loss": 0.5432, "step": 92 }, { "epoch": 0.03, "learning_rate": 9.965593118623726e-06, "loss": 0.5587, "step": 93 }, { "epoch": 0.03, "learning_rate": 9.965193038607723e-06, "loss": 0.3948, "step": 94 }, { "epoch": 0.03, "learning_rate": 9.96479295859172e-06, "loss": 0.4876, "step": 95 }, { "epoch": 0.03, "learning_rate": 9.964392878575716e-06, "loss": 0.6786, "step": 96 }, { "epoch": 0.03, "learning_rate": 9.963992798559713e-06, "loss": 0.4382, "step": 97 }, { "epoch": 0.03, "learning_rate": 9.96359271854371e-06, "loss": 0.5397, "step": 98 }, { "epoch": 0.03, "learning_rate": 9.963192638527706e-06, "loss": 0.362, "step": 99 }, { "epoch": 0.03, "learning_rate": 9.962792558511703e-06, "loss": 0.6364, "step": 100 }, { "epoch": 0.03, "learning_rate": 9.962392478495701e-06, "loss": 0.4415, "step": 101 }, { "epoch": 0.03, "learning_rate": 9.961992398479696e-06, "loss": 0.5632, "step": 102 }, { "epoch": 0.03, "learning_rate": 9.961592318463693e-06, "loss": 0.6013, "step": 103 }, { "epoch": 0.03, "learning_rate": 9.96119223844769e-06, "loss": 0.5648, "step": 104 }, { "epoch": 0.03, "learning_rate": 9.960792158431686e-06, "loss": 0.3298, "step": 105 }, { "epoch": 0.03, "learning_rate": 9.960392078415685e-06, "loss": 0.4127, "step": 106 }, { "epoch": 0.03, "learning_rate": 9.959991998399681e-06, "loss": 0.4534, "step": 107 }, { "epoch": 0.03, "learning_rate": 9.959591918383678e-06, "loss": 0.5458, "step": 108 }, { "epoch": 0.03, "learning_rate": 9.959191838367675e-06, "loss": 0.4436, "step": 109 }, { "epoch": 0.03, "learning_rate": 9.958791758351671e-06, "loss": 0.3407, "step": 110 }, { "epoch": 0.03, "learning_rate": 9.958391678335668e-06, "loss": 0.6319, "step": 111 }, { "epoch": 0.03, "learning_rate": 9.957991598319665e-06, "loss": 0.4963, "step": 112 }, { "epoch": 0.04, "learning_rate": 9.957591518303661e-06, "loss": 0.4868, "step": 113 }, { "epoch": 0.04, "learning_rate": 9.957191438287658e-06, "loss": 0.4327, "step": 114 }, { "epoch": 0.04, "learning_rate": 9.956791358271655e-06, "loss": 0.5688, "step": 115 }, { "epoch": 0.04, "learning_rate": 9.956391278255651e-06, "loss": 0.3451, "step": 116 }, { "epoch": 0.04, "learning_rate": 9.955991198239648e-06, "loss": 0.7755, "step": 117 }, { "epoch": 0.04, "learning_rate": 9.955591118223645e-06, "loss": 0.5532, "step": 118 }, { "epoch": 0.04, "learning_rate": 9.955191038207642e-06, "loss": 0.4852, "step": 119 }, { "epoch": 0.04, "learning_rate": 9.95479095819164e-06, "loss": 0.7424, "step": 120 }, { "epoch": 0.04, "learning_rate": 9.954390878175637e-06, "loss": 0.3389, "step": 121 }, { "epoch": 0.04, "learning_rate": 9.953990798159633e-06, "loss": 0.6283, "step": 122 }, { "epoch": 0.04, "learning_rate": 9.95359071814363e-06, "loss": 0.4868, "step": 123 }, { "epoch": 0.04, "learning_rate": 9.953190638127627e-06, "loss": 0.6801, "step": 124 }, { "epoch": 0.04, "learning_rate": 9.952790558111623e-06, "loss": 0.3732, "step": 125 }, { "epoch": 0.04, "learning_rate": 9.95239047809562e-06, "loss": 0.3511, "step": 126 }, { "epoch": 0.04, "learning_rate": 9.951990398079617e-06, "loss": 0.6292, "step": 127 }, { "epoch": 0.04, "learning_rate": 9.951590318063613e-06, "loss": 0.3595, "step": 128 }, { "epoch": 0.04, "learning_rate": 9.95119023804761e-06, "loss": 0.4395, "step": 129 }, { "epoch": 0.04, "learning_rate": 9.950790158031607e-06, "loss": 0.5519, "step": 130 }, { "epoch": 0.04, "learning_rate": 9.950390078015603e-06, "loss": 0.4559, "step": 131 }, { "epoch": 0.04, "learning_rate": 9.9499899979996e-06, "loss": 0.5539, "step": 132 }, { "epoch": 0.04, "learning_rate": 9.949589917983597e-06, "loss": 0.5456, "step": 133 }, { "epoch": 0.04, "learning_rate": 9.949189837967595e-06, "loss": 0.3494, "step": 134 }, { "epoch": 0.04, "learning_rate": 9.948789757951592e-06, "loss": 0.4619, "step": 135 }, { "epoch": 0.04, "learning_rate": 9.948389677935589e-06, "loss": 0.5023, "step": 136 }, { "epoch": 0.04, "learning_rate": 9.947989597919585e-06, "loss": 0.4609, "step": 137 }, { "epoch": 0.04, "learning_rate": 9.947589517903582e-06, "loss": 0.4408, "step": 138 }, { "epoch": 0.04, "learning_rate": 9.947189437887579e-06, "loss": 0.4588, "step": 139 }, { "epoch": 0.04, "learning_rate": 9.946789357871575e-06, "loss": 0.3758, "step": 140 }, { "epoch": 0.04, "learning_rate": 9.946389277855572e-06, "loss": 0.5017, "step": 141 }, { "epoch": 0.04, "learning_rate": 9.945989197839569e-06, "loss": 0.5298, "step": 142 }, { "epoch": 0.04, "learning_rate": 9.945589117823565e-06, "loss": 0.4785, "step": 143 }, { "epoch": 0.04, "learning_rate": 9.945189037807562e-06, "loss": 0.2794, "step": 144 }, { "epoch": 0.05, "learning_rate": 9.944788957791559e-06, "loss": 0.6405, "step": 145 }, { "epoch": 0.05, "learning_rate": 9.944388877775555e-06, "loss": 0.4634, "step": 146 }, { "epoch": 0.05, "learning_rate": 9.943988797759552e-06, "loss": 0.3557, "step": 147 }, { "epoch": 0.05, "learning_rate": 9.94358871774355e-06, "loss": 0.413, "step": 148 }, { "epoch": 0.05, "learning_rate": 9.943188637727547e-06, "loss": 0.5378, "step": 149 }, { "epoch": 0.05, "learning_rate": 9.942788557711544e-06, "loss": 0.4625, "step": 150 }, { "epoch": 0.05, "learning_rate": 9.94238847769554e-06, "loss": 0.3033, "step": 151 }, { "epoch": 0.05, "learning_rate": 9.941988397679537e-06, "loss": 0.3511, "step": 152 }, { "epoch": 0.05, "learning_rate": 9.941588317663532e-06, "loss": 0.359, "step": 153 }, { "epoch": 0.05, "learning_rate": 9.94118823764753e-06, "loss": 0.4354, "step": 154 }, { "epoch": 0.05, "learning_rate": 9.940788157631527e-06, "loss": 0.3314, "step": 155 }, { "epoch": 0.05, "learning_rate": 9.940388077615524e-06, "loss": 0.4444, "step": 156 }, { "epoch": 0.05, "learning_rate": 9.93998799759952e-06, "loss": 0.4215, "step": 157 }, { "epoch": 0.05, "learning_rate": 9.939587917583517e-06, "loss": 0.3707, "step": 158 }, { "epoch": 0.05, "learning_rate": 9.939187837567514e-06, "loss": 0.413, "step": 159 }, { "epoch": 0.05, "learning_rate": 9.93878775755151e-06, "loss": 0.3725, "step": 160 }, { "epoch": 0.05, "learning_rate": 9.938387677535507e-06, "loss": 0.5448, "step": 161 }, { "epoch": 0.05, "learning_rate": 9.937987597519506e-06, "loss": 0.492, "step": 162 }, { "epoch": 0.05, "learning_rate": 9.937587517503503e-06, "loss": 0.4456, "step": 163 }, { "epoch": 0.05, "learning_rate": 9.9371874374875e-06, "loss": 0.5179, "step": 164 }, { "epoch": 0.05, "learning_rate": 9.936787357471494e-06, "loss": 0.4283, "step": 165 }, { "epoch": 0.05, "learning_rate": 9.936387277455491e-06, "loss": 0.2836, "step": 166 }, { "epoch": 0.05, "learning_rate": 9.935987197439488e-06, "loss": 0.5283, "step": 167 }, { "epoch": 0.05, "learning_rate": 9.935587117423486e-06, "loss": 0.3726, "step": 168 }, { "epoch": 0.05, "learning_rate": 9.935187037407483e-06, "loss": 0.468, "step": 169 }, { "epoch": 0.05, "learning_rate": 9.93478695739148e-06, "loss": 0.7478, "step": 170 }, { "epoch": 0.05, "learning_rate": 9.934386877375476e-06, "loss": 0.4697, "step": 171 }, { "epoch": 0.05, "learning_rate": 9.933986797359473e-06, "loss": 0.4472, "step": 172 }, { "epoch": 0.05, "learning_rate": 9.93358671734347e-06, "loss": 0.4645, "step": 173 }, { "epoch": 0.05, "learning_rate": 9.933186637327466e-06, "loss": 0.5396, "step": 174 }, { "epoch": 0.05, "learning_rate": 9.932786557311463e-06, "loss": 0.375, "step": 175 }, { "epoch": 0.05, "learning_rate": 9.932386477295461e-06, "loss": 0.4189, "step": 176 }, { "epoch": 0.05, "learning_rate": 9.931986397279458e-06, "loss": 0.3426, "step": 177 }, { "epoch": 0.06, "learning_rate": 9.931586317263453e-06, "loss": 0.3226, "step": 178 }, { "epoch": 0.06, "learning_rate": 9.93118623724745e-06, "loss": 0.4231, "step": 179 }, { "epoch": 0.06, "learning_rate": 9.930786157231446e-06, "loss": 0.4156, "step": 180 }, { "epoch": 0.06, "learning_rate": 9.930386077215443e-06, "loss": 0.3762, "step": 181 }, { "epoch": 0.06, "learning_rate": 9.929985997199441e-06, "loss": 0.4996, "step": 182 }, { "epoch": 0.06, "learning_rate": 9.929585917183438e-06, "loss": 0.3647, "step": 183 }, { "epoch": 0.06, "learning_rate": 9.929185837167435e-06, "loss": 0.3953, "step": 184 }, { "epoch": 0.06, "learning_rate": 9.928785757151431e-06, "loss": 0.4543, "step": 185 }, { "epoch": 0.06, "learning_rate": 9.928385677135428e-06, "loss": 0.3987, "step": 186 }, { "epoch": 0.06, "learning_rate": 9.927985597119425e-06, "loss": 0.7736, "step": 187 }, { "epoch": 0.06, "learning_rate": 9.927585517103421e-06, "loss": 0.4186, "step": 188 }, { "epoch": 0.06, "learning_rate": 9.927185437087418e-06, "loss": 0.3986, "step": 189 }, { "epoch": 0.06, "learning_rate": 9.926785357071415e-06, "loss": 0.5774, "step": 190 }, { "epoch": 0.06, "learning_rate": 9.926385277055411e-06, "loss": 0.5038, "step": 191 }, { "epoch": 0.06, "learning_rate": 9.925985197039408e-06, "loss": 0.394, "step": 192 }, { "epoch": 0.06, "learning_rate": 9.925585117023405e-06, "loss": 0.2994, "step": 193 }, { "epoch": 0.06, "learning_rate": 9.925185037007401e-06, "loss": 0.3516, "step": 194 }, { "epoch": 0.06, "learning_rate": 9.924784956991398e-06, "loss": 0.5172, "step": 195 }, { "epoch": 0.06, "learning_rate": 9.924384876975397e-06, "loss": 0.5639, "step": 196 }, { "epoch": 0.06, "learning_rate": 9.923984796959393e-06, "loss": 0.5086, "step": 197 }, { "epoch": 0.06, "learning_rate": 9.92358471694339e-06, "loss": 0.4965, "step": 198 }, { "epoch": 0.06, "learning_rate": 9.923184636927387e-06, "loss": 0.7063, "step": 199 }, { "epoch": 0.06, "learning_rate": 9.922784556911383e-06, "loss": 0.5526, "step": 200 }, { "epoch": 0.06, "learning_rate": 9.92238447689538e-06, "loss": 0.6607, "step": 201 }, { "epoch": 0.06, "learning_rate": 9.921984396879377e-06, "loss": 0.4356, "step": 202 }, { "epoch": 0.06, "learning_rate": 9.921584316863373e-06, "loss": 0.4033, "step": 203 }, { "epoch": 0.06, "learning_rate": 9.92118423684737e-06, "loss": 0.5142, "step": 204 }, { "epoch": 0.06, "learning_rate": 9.920784156831367e-06, "loss": 0.4176, "step": 205 }, { "epoch": 0.06, "learning_rate": 9.920384076815363e-06, "loss": 0.4428, "step": 206 }, { "epoch": 0.06, "learning_rate": 9.91998399679936e-06, "loss": 0.4636, "step": 207 }, { "epoch": 0.06, "learning_rate": 9.919583916783357e-06, "loss": 0.3924, "step": 208 }, { "epoch": 0.06, "learning_rate": 9.919183836767353e-06, "loss": 0.4299, "step": 209 }, { "epoch": 0.07, "learning_rate": 9.918783756751352e-06, "loss": 0.6186, "step": 210 }, { "epoch": 0.07, "learning_rate": 9.918383676735349e-06, "loss": 0.4882, "step": 211 }, { "epoch": 0.07, "learning_rate": 9.917983596719345e-06, "loss": 0.402, "step": 212 }, { "epoch": 0.07, "learning_rate": 9.917583516703342e-06, "loss": 0.6252, "step": 213 }, { "epoch": 0.07, "learning_rate": 9.917183436687339e-06, "loss": 0.4689, "step": 214 }, { "epoch": 0.07, "learning_rate": 9.916783356671335e-06, "loss": 0.3799, "step": 215 }, { "epoch": 0.07, "learning_rate": 9.916383276655332e-06, "loss": 0.4896, "step": 216 }, { "epoch": 0.07, "learning_rate": 9.915983196639329e-06, "loss": 0.4032, "step": 217 }, { "epoch": 0.07, "learning_rate": 9.915583116623325e-06, "loss": 0.4169, "step": 218 }, { "epoch": 0.07, "learning_rate": 9.915183036607322e-06, "loss": 0.6688, "step": 219 }, { "epoch": 0.07, "learning_rate": 9.914782956591319e-06, "loss": 0.4468, "step": 220 }, { "epoch": 0.07, "learning_rate": 9.914382876575315e-06, "loss": 0.3615, "step": 221 }, { "epoch": 0.07, "learning_rate": 9.913982796559312e-06, "loss": 0.4039, "step": 222 }, { "epoch": 0.07, "learning_rate": 9.913582716543309e-06, "loss": 0.5037, "step": 223 }, { "epoch": 0.07, "learning_rate": 9.913182636527307e-06, "loss": 0.3216, "step": 224 }, { "epoch": 0.07, "learning_rate": 9.912782556511304e-06, "loss": 0.4708, "step": 225 }, { "epoch": 0.07, "learning_rate": 9.9123824764953e-06, "loss": 0.4228, "step": 226 }, { "epoch": 0.07, "learning_rate": 9.911982396479297e-06, "loss": 0.6063, "step": 227 }, { "epoch": 0.07, "learning_rate": 9.911582316463292e-06, "loss": 0.3692, "step": 228 }, { "epoch": 0.07, "learning_rate": 9.91118223644729e-06, "loss": 0.4614, "step": 229 }, { "epoch": 0.07, "learning_rate": 9.910782156431287e-06, "loss": 0.4602, "step": 230 }, { "epoch": 0.07, "learning_rate": 9.910382076415284e-06, "loss": 0.3787, "step": 231 }, { "epoch": 0.07, "learning_rate": 9.90998199639928e-06, "loss": 0.4611, "step": 232 }, { "epoch": 0.07, "learning_rate": 9.909581916383277e-06, "loss": 0.3875, "step": 233 }, { "epoch": 0.07, "learning_rate": 9.909181836367274e-06, "loss": 0.5002, "step": 234 }, { "epoch": 0.07, "learning_rate": 9.90878175635127e-06, "loss": 0.3856, "step": 235 }, { "epoch": 0.07, "learning_rate": 9.908381676335267e-06, "loss": 0.4107, "step": 236 }, { "epoch": 0.07, "learning_rate": 9.907981596319266e-06, "loss": 0.4246, "step": 237 }, { "epoch": 0.07, "learning_rate": 9.907581516303262e-06, "loss": 0.4691, "step": 238 }, { "epoch": 0.07, "learning_rate": 9.907181436287259e-06, "loss": 0.5272, "step": 239 }, { "epoch": 0.07, "learning_rate": 9.906781356271256e-06, "loss": 0.4217, "step": 240 }, { "epoch": 0.07, "learning_rate": 9.90638127625525e-06, "loss": 0.5108, "step": 241 }, { "epoch": 0.08, "learning_rate": 9.905981196239248e-06, "loss": 0.471, "step": 242 }, { "epoch": 0.08, "learning_rate": 9.905581116223246e-06, "loss": 0.3909, "step": 243 }, { "epoch": 0.08, "learning_rate": 9.905181036207243e-06, "loss": 0.616, "step": 244 }, { "epoch": 0.08, "learning_rate": 9.90478095619124e-06, "loss": 0.3647, "step": 245 }, { "epoch": 0.08, "learning_rate": 9.904380876175236e-06, "loss": 0.5528, "step": 246 }, { "epoch": 0.08, "learning_rate": 9.903980796159233e-06, "loss": 0.3266, "step": 247 }, { "epoch": 0.08, "learning_rate": 9.90358071614323e-06, "loss": 0.3582, "step": 248 }, { "epoch": 0.08, "learning_rate": 9.903180636127226e-06, "loss": 0.3762, "step": 249 }, { "epoch": 0.08, "learning_rate": 9.902780556111223e-06, "loss": 0.4799, "step": 250 }, { "epoch": 0.08, "learning_rate": 9.902380476095221e-06, "loss": 0.3608, "step": 251 }, { "epoch": 0.08, "learning_rate": 9.901980396079218e-06, "loss": 0.5127, "step": 252 }, { "epoch": 0.08, "learning_rate": 9.901580316063213e-06, "loss": 0.4957, "step": 253 }, { "epoch": 0.08, "learning_rate": 9.90118023604721e-06, "loss": 0.3908, "step": 254 }, { "epoch": 0.08, "learning_rate": 9.900780156031206e-06, "loss": 0.4733, "step": 255 }, { "epoch": 0.08, "learning_rate": 9.900380076015203e-06, "loss": 0.1914, "step": 256 }, { "epoch": 0.08, "learning_rate": 9.899979995999201e-06, "loss": 0.4059, "step": 257 }, { "epoch": 0.08, "learning_rate": 9.899579915983198e-06, "loss": 0.6075, "step": 258 }, { "epoch": 0.08, "learning_rate": 9.899179835967195e-06, "loss": 0.4719, "step": 259 }, { "epoch": 0.08, "learning_rate": 9.898779755951191e-06, "loss": 0.4474, "step": 260 }, { "epoch": 0.08, "learning_rate": 9.898379675935188e-06, "loss": 0.6023, "step": 261 }, { "epoch": 0.08, "learning_rate": 9.897979595919185e-06, "loss": 0.4246, "step": 262 }, { "epoch": 0.08, "learning_rate": 9.897579515903181e-06, "loss": 0.4539, "step": 263 }, { "epoch": 0.08, "learning_rate": 9.897179435887178e-06, "loss": 0.3097, "step": 264 }, { "epoch": 0.08, "learning_rate": 9.896779355871176e-06, "loss": 0.3023, "step": 265 }, { "epoch": 0.08, "learning_rate": 9.896379275855171e-06, "loss": 0.3691, "step": 266 }, { "epoch": 0.08, "learning_rate": 9.895979195839168e-06, "loss": 0.3188, "step": 267 }, { "epoch": 0.08, "learning_rate": 9.895579115823165e-06, "loss": 0.4747, "step": 268 }, { "epoch": 0.08, "learning_rate": 9.895179035807161e-06, "loss": 0.3419, "step": 269 }, { "epoch": 0.08, "learning_rate": 9.894778955791158e-06, "loss": 0.4088, "step": 270 }, { "epoch": 0.08, "learning_rate": 9.894378875775156e-06, "loss": 0.3957, "step": 271 }, { "epoch": 0.08, "learning_rate": 9.893978795759153e-06, "loss": 0.3839, "step": 272 }, { "epoch": 0.08, "learning_rate": 9.89357871574315e-06, "loss": 0.3902, "step": 273 }, { "epoch": 0.09, "learning_rate": 9.893178635727147e-06, "loss": 0.4766, "step": 274 }, { "epoch": 0.09, "learning_rate": 9.892778555711143e-06, "loss": 0.4492, "step": 275 }, { "epoch": 0.09, "learning_rate": 9.89237847569514e-06, "loss": 0.5188, "step": 276 }, { "epoch": 0.09, "learning_rate": 9.891978395679137e-06, "loss": 0.4029, "step": 277 }, { "epoch": 0.09, "learning_rate": 9.891578315663133e-06, "loss": 0.4823, "step": 278 }, { "epoch": 0.09, "learning_rate": 9.89117823564713e-06, "loss": 0.6143, "step": 279 }, { "epoch": 0.09, "learning_rate": 9.890778155631127e-06, "loss": 0.503, "step": 280 }, { "epoch": 0.09, "learning_rate": 9.890378075615123e-06, "loss": 0.3372, "step": 281 }, { "epoch": 0.09, "learning_rate": 9.88997799559912e-06, "loss": 0.4913, "step": 282 }, { "epoch": 0.09, "learning_rate": 9.889577915583117e-06, "loss": 0.4241, "step": 283 }, { "epoch": 0.09, "learning_rate": 9.889177835567113e-06, "loss": 0.6567, "step": 284 }, { "epoch": 0.09, "learning_rate": 9.888777755551112e-06, "loss": 0.3255, "step": 285 }, { "epoch": 0.09, "learning_rate": 9.888377675535108e-06, "loss": 0.3684, "step": 286 }, { "epoch": 0.09, "learning_rate": 9.887977595519105e-06, "loss": 0.4048, "step": 287 }, { "epoch": 0.09, "learning_rate": 9.887577515503102e-06, "loss": 0.3641, "step": 288 }, { "epoch": 0.09, "learning_rate": 9.887177435487099e-06, "loss": 0.4247, "step": 289 }, { "epoch": 0.09, "learning_rate": 9.886777355471095e-06, "loss": 0.4465, "step": 290 }, { "epoch": 0.09, "learning_rate": 9.886377275455092e-06, "loss": 0.6473, "step": 291 }, { "epoch": 0.09, "learning_rate": 9.885977195439089e-06, "loss": 0.3489, "step": 292 }, { "epoch": 0.09, "learning_rate": 9.885577115423085e-06, "loss": 0.41, "step": 293 }, { "epoch": 0.09, "learning_rate": 9.885177035407082e-06, "loss": 0.5398, "step": 294 }, { "epoch": 0.09, "learning_rate": 9.884776955391079e-06, "loss": 0.315, "step": 295 }, { "epoch": 0.09, "learning_rate": 9.884376875375075e-06, "loss": 0.275, "step": 296 }, { "epoch": 0.09, "learning_rate": 9.883976795359072e-06, "loss": 0.3693, "step": 297 }, { "epoch": 0.09, "learning_rate": 9.883576715343069e-06, "loss": 0.2498, "step": 298 }, { "epoch": 0.09, "learning_rate": 9.883176635327067e-06, "loss": 0.3097, "step": 299 }, { "epoch": 0.09, "learning_rate": 9.882776555311064e-06, "loss": 0.3821, "step": 300 }, { "epoch": 0.09, "learning_rate": 9.88237647529506e-06, "loss": 0.6335, "step": 301 }, { "epoch": 0.09, "learning_rate": 9.881976395279057e-06, "loss": 0.4909, "step": 302 }, { "epoch": 0.09, "learning_rate": 9.881576315263054e-06, "loss": 0.3609, "step": 303 }, { "epoch": 0.09, "learning_rate": 9.88117623524705e-06, "loss": 0.3188, "step": 304 }, { "epoch": 0.09, "learning_rate": 9.880776155231047e-06, "loss": 0.4396, "step": 305 }, { "epoch": 0.09, "learning_rate": 9.880376075215044e-06, "loss": 0.6075, "step": 306 }, { "epoch": 0.1, "learning_rate": 9.87997599519904e-06, "loss": 0.4999, "step": 307 }, { "epoch": 0.1, "learning_rate": 9.879575915183037e-06, "loss": 0.4834, "step": 308 }, { "epoch": 0.1, "learning_rate": 9.879175835167034e-06, "loss": 0.448, "step": 309 }, { "epoch": 0.1, "learning_rate": 9.87877575515103e-06, "loss": 0.6293, "step": 310 }, { "epoch": 0.1, "learning_rate": 9.878375675135027e-06, "loss": 0.3337, "step": 311 }, { "epoch": 0.1, "learning_rate": 9.877975595119024e-06, "loss": 0.5648, "step": 312 }, { "epoch": 0.1, "learning_rate": 9.877575515103022e-06, "loss": 0.2719, "step": 313 }, { "epoch": 0.1, "learning_rate": 9.877175435087019e-06, "loss": 0.3962, "step": 314 }, { "epoch": 0.1, "learning_rate": 9.876775355071016e-06, "loss": 0.4938, "step": 315 }, { "epoch": 0.1, "learning_rate": 9.87637527505501e-06, "loss": 0.4303, "step": 316 }, { "epoch": 0.1, "learning_rate": 9.875975195039007e-06, "loss": 0.4929, "step": 317 }, { "epoch": 0.1, "learning_rate": 9.875575115023006e-06, "loss": 0.2997, "step": 318 }, { "epoch": 0.1, "learning_rate": 9.875175035007003e-06, "loss": 0.5006, "step": 319 }, { "epoch": 0.1, "learning_rate": 9.874774954991e-06, "loss": 0.425, "step": 320 }, { "epoch": 0.1, "learning_rate": 9.874374874974996e-06, "loss": 0.516, "step": 321 }, { "epoch": 0.1, "learning_rate": 9.873974794958993e-06, "loss": 0.3585, "step": 322 }, { "epoch": 0.1, "learning_rate": 9.87357471494299e-06, "loss": 0.4263, "step": 323 }, { "epoch": 0.1, "learning_rate": 9.873174634926986e-06, "loss": 0.3405, "step": 324 }, { "epoch": 0.1, "learning_rate": 9.872774554910983e-06, "loss": 0.3955, "step": 325 }, { "epoch": 0.1, "learning_rate": 9.87237447489498e-06, "loss": 0.3996, "step": 326 }, { "epoch": 0.1, "learning_rate": 9.871974394878978e-06, "loss": 0.4475, "step": 327 }, { "epoch": 0.1, "learning_rate": 9.871574314862974e-06, "loss": 0.4245, "step": 328 }, { "epoch": 0.1, "learning_rate": 9.87117423484697e-06, "loss": 0.3217, "step": 329 }, { "epoch": 0.1, "learning_rate": 9.870774154830966e-06, "loss": 0.3528, "step": 330 }, { "epoch": 0.1, "learning_rate": 9.870374074814963e-06, "loss": 0.4775, "step": 331 }, { "epoch": 0.1, "learning_rate": 9.869973994798961e-06, "loss": 0.4186, "step": 332 }, { "epoch": 0.1, "learning_rate": 9.869573914782958e-06, "loss": 0.3984, "step": 333 }, { "epoch": 0.1, "learning_rate": 9.869173834766954e-06, "loss": 0.4254, "step": 334 }, { "epoch": 0.1, "learning_rate": 9.868773754750951e-06, "loss": 0.3615, "step": 335 }, { "epoch": 0.1, "learning_rate": 9.868373674734948e-06, "loss": 0.3395, "step": 336 }, { "epoch": 0.1, "learning_rate": 9.867973594718945e-06, "loss": 0.3696, "step": 337 }, { "epoch": 0.1, "learning_rate": 9.867573514702941e-06, "loss": 0.5443, "step": 338 }, { "epoch": 0.11, "learning_rate": 9.867173434686938e-06, "loss": 0.4456, "step": 339 }, { "epoch": 0.11, "learning_rate": 9.866773354670935e-06, "loss": 0.4587, "step": 340 }, { "epoch": 0.11, "learning_rate": 9.866373274654933e-06, "loss": 0.3813, "step": 341 }, { "epoch": 0.11, "learning_rate": 9.865973194638928e-06, "loss": 0.4041, "step": 342 }, { "epoch": 0.11, "learning_rate": 9.865573114622925e-06, "loss": 0.4676, "step": 343 }, { "epoch": 0.11, "learning_rate": 9.865173034606921e-06, "loss": 0.3315, "step": 344 }, { "epoch": 0.11, "learning_rate": 9.864772954590918e-06, "loss": 0.2987, "step": 345 }, { "epoch": 0.11, "learning_rate": 9.864372874574916e-06, "loss": 0.3725, "step": 346 }, { "epoch": 0.11, "learning_rate": 9.863972794558913e-06, "loss": 0.5553, "step": 347 }, { "epoch": 0.11, "learning_rate": 9.86357271454291e-06, "loss": 0.5732, "step": 348 }, { "epoch": 0.11, "learning_rate": 9.863172634526906e-06, "loss": 0.5056, "step": 349 }, { "epoch": 0.11, "learning_rate": 9.862772554510903e-06, "loss": 0.4051, "step": 350 }, { "epoch": 0.11, "learning_rate": 9.8623724744949e-06, "loss": 0.5743, "step": 351 }, { "epoch": 0.11, "learning_rate": 9.861972394478897e-06, "loss": 0.3769, "step": 352 }, { "epoch": 0.11, "learning_rate": 9.861572314462893e-06, "loss": 0.5295, "step": 353 }, { "epoch": 0.11, "learning_rate": 9.86117223444689e-06, "loss": 0.3704, "step": 354 }, { "epoch": 0.11, "learning_rate": 9.860772154430887e-06, "loss": 0.5783, "step": 355 }, { "epoch": 0.11, "learning_rate": 9.860372074414883e-06, "loss": 0.3839, "step": 356 }, { "epoch": 0.11, "learning_rate": 9.85997199439888e-06, "loss": 0.4407, "step": 357 }, { "epoch": 0.11, "learning_rate": 9.859571914382877e-06, "loss": 0.2156, "step": 358 }, { "epoch": 0.11, "learning_rate": 9.859171834366873e-06, "loss": 0.3508, "step": 359 }, { "epoch": 0.11, "learning_rate": 9.858771754350872e-06, "loss": 0.3894, "step": 360 }, { "epoch": 0.11, "learning_rate": 9.858371674334868e-06, "loss": 0.4049, "step": 361 }, { "epoch": 0.11, "learning_rate": 9.857971594318865e-06, "loss": 0.3378, "step": 362 }, { "epoch": 0.11, "learning_rate": 9.857571514302862e-06, "loss": 0.4308, "step": 363 }, { "epoch": 0.11, "learning_rate": 9.857171434286858e-06, "loss": 0.4241, "step": 364 }, { "epoch": 0.11, "learning_rate": 9.856771354270855e-06, "loss": 0.424, "step": 365 }, { "epoch": 0.11, "learning_rate": 9.856371274254852e-06, "loss": 0.4294, "step": 366 }, { "epoch": 0.11, "learning_rate": 9.855971194238849e-06, "loss": 0.4119, "step": 367 }, { "epoch": 0.11, "learning_rate": 9.855571114222845e-06, "loss": 0.3959, "step": 368 }, { "epoch": 0.11, "learning_rate": 9.855171034206842e-06, "loss": 0.3898, "step": 369 }, { "epoch": 0.11, "learning_rate": 9.854770954190839e-06, "loss": 0.4159, "step": 370 }, { "epoch": 0.12, "learning_rate": 9.854370874174835e-06, "loss": 0.7481, "step": 371 }, { "epoch": 0.12, "learning_rate": 9.853970794158832e-06, "loss": 0.4017, "step": 372 }, { "epoch": 0.12, "learning_rate": 9.853570714142829e-06, "loss": 0.4677, "step": 373 }, { "epoch": 0.12, "learning_rate": 9.853170634126827e-06, "loss": 0.4603, "step": 374 }, { "epoch": 0.12, "learning_rate": 9.852770554110824e-06, "loss": 0.4413, "step": 375 }, { "epoch": 0.12, "learning_rate": 9.85237047409482e-06, "loss": 0.3747, "step": 376 }, { "epoch": 0.12, "learning_rate": 9.851970394078817e-06, "loss": 0.372, "step": 377 }, { "epoch": 0.12, "learning_rate": 9.851570314062814e-06, "loss": 0.3658, "step": 378 }, { "epoch": 0.12, "learning_rate": 9.85117023404681e-06, "loss": 0.246, "step": 379 }, { "epoch": 0.12, "learning_rate": 9.850770154030807e-06, "loss": 0.4937, "step": 380 }, { "epoch": 0.12, "learning_rate": 9.850370074014804e-06, "loss": 0.4274, "step": 381 }, { "epoch": 0.12, "learning_rate": 9.8499699939988e-06, "loss": 0.3927, "step": 382 }, { "epoch": 0.12, "learning_rate": 9.849569913982797e-06, "loss": 0.4583, "step": 383 }, { "epoch": 0.12, "learning_rate": 9.849169833966794e-06, "loss": 0.3356, "step": 384 }, { "epoch": 0.12, "learning_rate": 9.84876975395079e-06, "loss": 0.3452, "step": 385 }, { "epoch": 0.12, "learning_rate": 9.848369673934787e-06, "loss": 0.3304, "step": 386 }, { "epoch": 0.12, "learning_rate": 9.847969593918784e-06, "loss": 0.3702, "step": 387 }, { "epoch": 0.12, "learning_rate": 9.847569513902782e-06, "loss": 0.5113, "step": 388 }, { "epoch": 0.12, "learning_rate": 9.847169433886779e-06, "loss": 0.3195, "step": 389 }, { "epoch": 0.12, "learning_rate": 9.846769353870776e-06, "loss": 0.3027, "step": 390 }, { "epoch": 0.12, "learning_rate": 9.846369273854772e-06, "loss": 0.2996, "step": 391 }, { "epoch": 0.12, "learning_rate": 9.845969193838767e-06, "loss": 0.364, "step": 392 }, { "epoch": 0.12, "learning_rate": 9.845569113822766e-06, "loss": 0.4159, "step": 393 }, { "epoch": 0.12, "learning_rate": 9.845169033806762e-06, "loss": 0.478, "step": 394 }, { "epoch": 0.12, "learning_rate": 9.844768953790759e-06, "loss": 0.2684, "step": 395 }, { "epoch": 0.12, "learning_rate": 9.844368873774756e-06, "loss": 0.3953, "step": 396 }, { "epoch": 0.12, "learning_rate": 9.843968793758753e-06, "loss": 0.5044, "step": 397 }, { "epoch": 0.12, "learning_rate": 9.84356871374275e-06, "loss": 0.6002, "step": 398 }, { "epoch": 0.12, "learning_rate": 9.843168633726746e-06, "loss": 0.4507, "step": 399 }, { "epoch": 0.12, "learning_rate": 9.842768553710743e-06, "loss": 0.3376, "step": 400 }, { "epoch": 0.12, "learning_rate": 9.84236847369474e-06, "loss": 0.4124, "step": 401 }, { "epoch": 0.12, "learning_rate": 9.841968393678738e-06, "loss": 0.3544, "step": 402 }, { "epoch": 0.13, "learning_rate": 9.841568313662734e-06, "loss": 0.4907, "step": 403 }, { "epoch": 0.13, "learning_rate": 9.841168233646731e-06, "loss": 0.4675, "step": 404 }, { "epoch": 0.13, "learning_rate": 9.840768153630726e-06, "loss": 0.4146, "step": 405 }, { "epoch": 0.13, "learning_rate": 9.840368073614723e-06, "loss": 0.3994, "step": 406 }, { "epoch": 0.13, "learning_rate": 9.839967993598721e-06, "loss": 0.2793, "step": 407 }, { "epoch": 0.13, "learning_rate": 9.839567913582718e-06, "loss": 0.5492, "step": 408 }, { "epoch": 0.13, "learning_rate": 9.839167833566714e-06, "loss": 0.2658, "step": 409 }, { "epoch": 0.13, "learning_rate": 9.838767753550711e-06, "loss": 0.3657, "step": 410 }, { "epoch": 0.13, "learning_rate": 9.838367673534708e-06, "loss": 0.396, "step": 411 }, { "epoch": 0.13, "learning_rate": 9.837967593518704e-06, "loss": 0.309, "step": 412 }, { "epoch": 0.13, "learning_rate": 9.837567513502701e-06, "loss": 0.3392, "step": 413 }, { "epoch": 0.13, "learning_rate": 9.837167433486698e-06, "loss": 0.3922, "step": 414 }, { "epoch": 0.13, "learning_rate": 9.836767353470695e-06, "loss": 0.4359, "step": 415 }, { "epoch": 0.13, "learning_rate": 9.836367273454693e-06, "loss": 0.4371, "step": 416 }, { "epoch": 0.13, "learning_rate": 9.835967193438688e-06, "loss": 0.4974, "step": 417 }, { "epoch": 0.13, "learning_rate": 9.835567113422685e-06, "loss": 0.3638, "step": 418 }, { "epoch": 0.13, "learning_rate": 9.835167033406681e-06, "loss": 0.4898, "step": 419 }, { "epoch": 0.13, "learning_rate": 9.834766953390678e-06, "loss": 0.2648, "step": 420 }, { "epoch": 0.13, "learning_rate": 9.834366873374676e-06, "loss": 0.3475, "step": 421 }, { "epoch": 0.13, "learning_rate": 9.833966793358673e-06, "loss": 0.3729, "step": 422 }, { "epoch": 0.13, "learning_rate": 9.83356671334267e-06, "loss": 0.2557, "step": 423 }, { "epoch": 0.13, "learning_rate": 9.833166633326666e-06, "loss": 0.5903, "step": 424 }, { "epoch": 0.13, "learning_rate": 9.832766553310663e-06, "loss": 0.1888, "step": 425 }, { "epoch": 0.13, "learning_rate": 9.83236647329466e-06, "loss": 0.4745, "step": 426 }, { "epoch": 0.13, "learning_rate": 9.831966393278656e-06, "loss": 0.4, "step": 427 }, { "epoch": 0.13, "learning_rate": 9.831566313262653e-06, "loss": 0.3354, "step": 428 }, { "epoch": 0.13, "learning_rate": 9.83116623324665e-06, "loss": 0.3676, "step": 429 }, { "epoch": 0.13, "learning_rate": 9.830766153230647e-06, "loss": 0.2752, "step": 430 }, { "epoch": 0.13, "learning_rate": 9.830366073214643e-06, "loss": 0.4867, "step": 431 }, { "epoch": 0.13, "learning_rate": 9.82996599319864e-06, "loss": 0.4848, "step": 432 }, { "epoch": 0.13, "learning_rate": 9.829565913182637e-06, "loss": 0.2639, "step": 433 }, { "epoch": 0.13, "learning_rate": 9.829165833166633e-06, "loss": 0.6031, "step": 434 }, { "epoch": 0.14, "learning_rate": 9.828765753150632e-06, "loss": 0.487, "step": 435 }, { "epoch": 0.14, "learning_rate": 9.828365673134628e-06, "loss": 0.2789, "step": 436 }, { "epoch": 0.14, "learning_rate": 9.827965593118625e-06, "loss": 0.3909, "step": 437 }, { "epoch": 0.14, "learning_rate": 9.827565513102622e-06, "loss": 0.4277, "step": 438 }, { "epoch": 0.14, "learning_rate": 9.827165433086618e-06, "loss": 0.4308, "step": 439 }, { "epoch": 0.14, "learning_rate": 9.826765353070615e-06, "loss": 0.3281, "step": 440 }, { "epoch": 0.14, "learning_rate": 9.826365273054612e-06, "loss": 0.517, "step": 441 }, { "epoch": 0.14, "learning_rate": 9.825965193038608e-06, "loss": 0.4265, "step": 442 }, { "epoch": 0.14, "learning_rate": 9.825565113022605e-06, "loss": 0.4272, "step": 443 }, { "epoch": 0.14, "learning_rate": 9.825165033006602e-06, "loss": 0.3961, "step": 444 }, { "epoch": 0.14, "learning_rate": 9.824764952990599e-06, "loss": 0.4328, "step": 445 }, { "epoch": 0.14, "learning_rate": 9.824364872974595e-06, "loss": 0.4167, "step": 446 }, { "epoch": 0.14, "learning_rate": 9.823964792958592e-06, "loss": 0.427, "step": 447 }, { "epoch": 0.14, "learning_rate": 9.823564712942589e-06, "loss": 0.4631, "step": 448 }, { "epoch": 0.14, "learning_rate": 9.823164632926587e-06, "loss": 0.4002, "step": 449 }, { "epoch": 0.14, "learning_rate": 9.822764552910584e-06, "loss": 0.4549, "step": 450 }, { "epoch": 0.14, "learning_rate": 9.82236447289458e-06, "loss": 0.3771, "step": 451 }, { "epoch": 0.14, "learning_rate": 9.821964392878577e-06, "loss": 0.3849, "step": 452 }, { "epoch": 0.14, "learning_rate": 9.821564312862574e-06, "loss": 0.265, "step": 453 }, { "epoch": 0.14, "learning_rate": 9.82116423284657e-06, "loss": 0.4684, "step": 454 }, { "epoch": 0.14, "learning_rate": 9.820764152830567e-06, "loss": 0.3322, "step": 455 }, { "epoch": 0.14, "learning_rate": 9.820364072814564e-06, "loss": 0.5235, "step": 456 }, { "epoch": 0.14, "learning_rate": 9.81996399279856e-06, "loss": 0.4495, "step": 457 }, { "epoch": 0.14, "learning_rate": 9.819563912782557e-06, "loss": 0.6831, "step": 458 }, { "epoch": 0.14, "learning_rate": 9.819163832766554e-06, "loss": 0.3378, "step": 459 }, { "epoch": 0.14, "learning_rate": 9.81876375275055e-06, "loss": 0.561, "step": 460 }, { "epoch": 0.14, "learning_rate": 9.818363672734547e-06, "loss": 0.3876, "step": 461 }, { "epoch": 0.14, "learning_rate": 9.817963592718544e-06, "loss": 0.5121, "step": 462 }, { "epoch": 0.14, "learning_rate": 9.817563512702542e-06, "loss": 0.3917, "step": 463 }, { "epoch": 0.14, "learning_rate": 9.817163432686539e-06, "loss": 0.222, "step": 464 }, { "epoch": 0.14, "learning_rate": 9.816763352670536e-06, "loss": 0.323, "step": 465 }, { "epoch": 0.14, "learning_rate": 9.816363272654532e-06, "loss": 0.3184, "step": 466 }, { "epoch": 0.14, "learning_rate": 9.815963192638529e-06, "loss": 0.3125, "step": 467 }, { "epoch": 0.15, "learning_rate": 9.815563112622524e-06, "loss": 0.3644, "step": 468 }, { "epoch": 0.15, "learning_rate": 9.815163032606522e-06, "loss": 0.395, "step": 469 }, { "epoch": 0.15, "learning_rate": 9.814762952590519e-06, "loss": 0.4587, "step": 470 }, { "epoch": 0.15, "learning_rate": 9.814362872574516e-06, "loss": 0.2824, "step": 471 }, { "epoch": 0.15, "learning_rate": 9.813962792558512e-06, "loss": 0.5093, "step": 472 }, { "epoch": 0.15, "learning_rate": 9.813562712542509e-06, "loss": 0.5277, "step": 473 }, { "epoch": 0.15, "learning_rate": 9.813162632526506e-06, "loss": 0.5615, "step": 474 }, { "epoch": 0.15, "learning_rate": 9.812762552510503e-06, "loss": 0.4272, "step": 475 }, { "epoch": 0.15, "learning_rate": 9.8123624724945e-06, "loss": 0.4131, "step": 476 }, { "epoch": 0.15, "learning_rate": 9.811962392478498e-06, "loss": 0.3628, "step": 477 }, { "epoch": 0.15, "learning_rate": 9.811562312462494e-06, "loss": 0.348, "step": 478 }, { "epoch": 0.15, "learning_rate": 9.811162232446491e-06, "loss": 0.4013, "step": 479 }, { "epoch": 0.15, "learning_rate": 9.810762152430486e-06, "loss": 0.5077, "step": 480 }, { "epoch": 0.15, "learning_rate": 9.810362072414483e-06, "loss": 0.2817, "step": 481 }, { "epoch": 0.15, "learning_rate": 9.80996199239848e-06, "loss": 0.2769, "step": 482 }, { "epoch": 0.15, "learning_rate": 9.809561912382478e-06, "loss": 0.3774, "step": 483 }, { "epoch": 0.15, "learning_rate": 9.809161832366474e-06, "loss": 0.4288, "step": 484 }, { "epoch": 0.15, "learning_rate": 9.808761752350471e-06, "loss": 0.3808, "step": 485 }, { "epoch": 0.15, "learning_rate": 9.808361672334468e-06, "loss": 0.4436, "step": 486 }, { "epoch": 0.15, "learning_rate": 9.807961592318464e-06, "loss": 0.5471, "step": 487 }, { "epoch": 0.15, "learning_rate": 9.807561512302461e-06, "loss": 0.5465, "step": 488 }, { "epoch": 0.15, "learning_rate": 9.807161432286458e-06, "loss": 0.3181, "step": 489 }, { "epoch": 0.15, "learning_rate": 9.806761352270454e-06, "loss": 0.2058, "step": 490 }, { "epoch": 0.15, "learning_rate": 9.806361272254453e-06, "loss": 0.4182, "step": 491 }, { "epoch": 0.15, "learning_rate": 9.80596119223845e-06, "loss": 0.3837, "step": 492 }, { "epoch": 0.15, "learning_rate": 9.805561112222445e-06, "loss": 0.4396, "step": 493 }, { "epoch": 0.15, "learning_rate": 9.805161032206441e-06, "loss": 0.6124, "step": 494 }, { "epoch": 0.15, "learning_rate": 9.804760952190438e-06, "loss": 0.4334, "step": 495 }, { "epoch": 0.15, "learning_rate": 9.804360872174435e-06, "loss": 0.5563, "step": 496 }, { "epoch": 0.15, "learning_rate": 9.803960792158433e-06, "loss": 0.3992, "step": 497 }, { "epoch": 0.15, "learning_rate": 9.80356071214243e-06, "loss": 0.3438, "step": 498 }, { "epoch": 0.15, "learning_rate": 9.803160632126426e-06, "loss": 0.3197, "step": 499 }, { "epoch": 0.16, "learning_rate": 9.802760552110423e-06, "loss": 0.3228, "step": 500 }, { "epoch": 0.16, "learning_rate": 9.80236047209442e-06, "loss": 0.3019, "step": 501 }, { "epoch": 0.16, "learning_rate": 9.801960392078416e-06, "loss": 0.3986, "step": 502 }, { "epoch": 0.16, "learning_rate": 9.801560312062413e-06, "loss": 0.4223, "step": 503 }, { "epoch": 0.16, "learning_rate": 9.80116023204641e-06, "loss": 0.4195, "step": 504 }, { "epoch": 0.16, "learning_rate": 9.800760152030408e-06, "loss": 0.3093, "step": 505 }, { "epoch": 0.16, "learning_rate": 9.800360072014403e-06, "loss": 0.4227, "step": 506 }, { "epoch": 0.16, "learning_rate": 9.7999599919984e-06, "loss": 0.4668, "step": 507 }, { "epoch": 0.16, "learning_rate": 9.799559911982397e-06, "loss": 0.4649, "step": 508 }, { "epoch": 0.16, "learning_rate": 9.799159831966393e-06, "loss": 0.4797, "step": 509 }, { "epoch": 0.16, "learning_rate": 9.79875975195039e-06, "loss": 0.4713, "step": 510 }, { "epoch": 0.16, "learning_rate": 9.798359671934388e-06, "loss": 0.4619, "step": 511 }, { "epoch": 0.16, "learning_rate": 9.797959591918385e-06, "loss": 0.4856, "step": 512 }, { "epoch": 0.16, "learning_rate": 9.797559511902382e-06, "loss": 0.3533, "step": 513 }, { "epoch": 0.16, "learning_rate": 9.797159431886378e-06, "loss": 0.173, "step": 514 }, { "epoch": 0.16, "learning_rate": 9.796759351870375e-06, "loss": 0.3204, "step": 515 }, { "epoch": 0.16, "learning_rate": 9.796359271854372e-06, "loss": 0.2731, "step": 516 }, { "epoch": 0.16, "learning_rate": 9.795959191838368e-06, "loss": 0.2843, "step": 517 }, { "epoch": 0.16, "learning_rate": 9.795559111822365e-06, "loss": 0.3794, "step": 518 }, { "epoch": 0.16, "learning_rate": 9.795159031806362e-06, "loss": 0.4316, "step": 519 }, { "epoch": 0.16, "learning_rate": 9.794758951790358e-06, "loss": 0.4603, "step": 520 }, { "epoch": 0.16, "learning_rate": 9.794358871774355e-06, "loss": 0.3454, "step": 521 }, { "epoch": 0.16, "learning_rate": 9.793958791758352e-06, "loss": 0.2447, "step": 522 }, { "epoch": 0.16, "learning_rate": 9.793558711742349e-06, "loss": 0.5649, "step": 523 }, { "epoch": 0.16, "learning_rate": 9.793158631726345e-06, "loss": 0.3302, "step": 524 }, { "epoch": 0.16, "learning_rate": 9.792758551710344e-06, "loss": 0.4097, "step": 525 }, { "epoch": 0.16, "learning_rate": 9.79235847169434e-06, "loss": 0.4566, "step": 526 }, { "epoch": 0.16, "learning_rate": 9.791958391678337e-06, "loss": 0.5222, "step": 527 }, { "epoch": 0.16, "learning_rate": 9.791558311662334e-06, "loss": 0.3942, "step": 528 }, { "epoch": 0.16, "learning_rate": 9.79115823164633e-06, "loss": 0.3028, "step": 529 }, { "epoch": 0.16, "learning_rate": 9.790758151630327e-06, "loss": 0.536, "step": 530 }, { "epoch": 0.16, "learning_rate": 9.790358071614324e-06, "loss": 0.3609, "step": 531 }, { "epoch": 0.17, "learning_rate": 9.78995799159832e-06, "loss": 0.5362, "step": 532 }, { "epoch": 0.17, "learning_rate": 9.789557911582317e-06, "loss": 0.4382, "step": 533 }, { "epoch": 0.17, "learning_rate": 9.789157831566314e-06, "loss": 0.28, "step": 534 }, { "epoch": 0.17, "learning_rate": 9.78875775155031e-06, "loss": 0.3886, "step": 535 }, { "epoch": 0.17, "learning_rate": 9.788357671534307e-06, "loss": 0.4178, "step": 536 }, { "epoch": 0.17, "learning_rate": 9.787957591518304e-06, "loss": 0.3529, "step": 537 }, { "epoch": 0.17, "learning_rate": 9.787557511502302e-06, "loss": 0.3549, "step": 538 }, { "epoch": 0.17, "learning_rate": 9.787157431486299e-06, "loss": 0.2996, "step": 539 }, { "epoch": 0.17, "learning_rate": 9.786757351470296e-06, "loss": 0.2286, "step": 540 }, { "epoch": 0.17, "learning_rate": 9.786357271454292e-06, "loss": 0.2907, "step": 541 }, { "epoch": 0.17, "learning_rate": 9.785957191438289e-06, "loss": 0.6385, "step": 542 }, { "epoch": 0.17, "learning_rate": 9.785557111422284e-06, "loss": 0.3915, "step": 543 }, { "epoch": 0.17, "learning_rate": 9.785157031406282e-06, "loss": 0.459, "step": 544 }, { "epoch": 0.17, "learning_rate": 9.784756951390279e-06, "loss": 0.4063, "step": 545 }, { "epoch": 0.17, "learning_rate": 9.784356871374276e-06, "loss": 0.3943, "step": 546 }, { "epoch": 0.17, "learning_rate": 9.783956791358272e-06, "loss": 0.3943, "step": 547 }, { "epoch": 0.17, "learning_rate": 9.783556711342269e-06, "loss": 0.2746, "step": 548 }, { "epoch": 0.17, "learning_rate": 9.783156631326266e-06, "loss": 0.3147, "step": 549 }, { "epoch": 0.17, "learning_rate": 9.782756551310262e-06, "loss": 0.2785, "step": 550 }, { "epoch": 0.17, "learning_rate": 9.782356471294259e-06, "loss": 0.3203, "step": 551 }, { "epoch": 0.17, "learning_rate": 9.781956391278258e-06, "loss": 0.3449, "step": 552 }, { "epoch": 0.17, "learning_rate": 9.781556311262254e-06, "loss": 0.2529, "step": 553 }, { "epoch": 0.17, "learning_rate": 9.781156231246251e-06, "loss": 0.4816, "step": 554 }, { "epoch": 0.17, "learning_rate": 9.780756151230248e-06, "loss": 0.384, "step": 555 }, { "epoch": 0.17, "learning_rate": 9.780356071214243e-06, "loss": 0.439, "step": 556 }, { "epoch": 0.17, "learning_rate": 9.77995599119824e-06, "loss": 0.3062, "step": 557 }, { "epoch": 0.17, "learning_rate": 9.779555911182238e-06, "loss": 0.3833, "step": 558 }, { "epoch": 0.17, "learning_rate": 9.779155831166234e-06, "loss": 0.2159, "step": 559 }, { "epoch": 0.17, "learning_rate": 9.778755751150231e-06, "loss": 0.4595, "step": 560 }, { "epoch": 0.17, "learning_rate": 9.778355671134228e-06, "loss": 0.3495, "step": 561 }, { "epoch": 0.17, "learning_rate": 9.777955591118224e-06, "loss": 0.2532, "step": 562 }, { "epoch": 0.17, "learning_rate": 9.777555511102221e-06, "loss": 0.3652, "step": 563 }, { "epoch": 0.18, "learning_rate": 9.777155431086218e-06, "loss": 0.223, "step": 564 }, { "epoch": 0.18, "learning_rate": 9.776755351070214e-06, "loss": 0.2609, "step": 565 }, { "epoch": 0.18, "learning_rate": 9.776355271054213e-06, "loss": 0.2695, "step": 566 }, { "epoch": 0.18, "learning_rate": 9.77595519103821e-06, "loss": 0.2944, "step": 567 }, { "epoch": 0.18, "learning_rate": 9.775555111022206e-06, "loss": 0.2835, "step": 568 }, { "epoch": 0.18, "learning_rate": 9.775155031006201e-06, "loss": 0.3227, "step": 569 }, { "epoch": 0.18, "learning_rate": 9.774754950990198e-06, "loss": 0.3069, "step": 570 }, { "epoch": 0.18, "learning_rate": 9.774354870974195e-06, "loss": 0.3255, "step": 571 }, { "epoch": 0.18, "learning_rate": 9.773954790958193e-06, "loss": 0.4102, "step": 572 }, { "epoch": 0.18, "learning_rate": 9.77355471094219e-06, "loss": 0.3197, "step": 573 }, { "epoch": 0.18, "learning_rate": 9.773154630926186e-06, "loss": 0.2644, "step": 574 }, { "epoch": 0.18, "learning_rate": 9.772754550910183e-06, "loss": 0.4629, "step": 575 }, { "epoch": 0.18, "learning_rate": 9.77235447089418e-06, "loss": 0.3236, "step": 576 }, { "epoch": 0.18, "learning_rate": 9.771954390878176e-06, "loss": 0.6201, "step": 577 }, { "epoch": 0.18, "learning_rate": 9.771554310862173e-06, "loss": 0.2394, "step": 578 }, { "epoch": 0.18, "learning_rate": 9.77115423084617e-06, "loss": 0.3889, "step": 579 }, { "epoch": 0.18, "learning_rate": 9.770754150830168e-06, "loss": 0.3927, "step": 580 }, { "epoch": 0.18, "learning_rate": 9.770354070814163e-06, "loss": 0.3099, "step": 581 }, { "epoch": 0.18, "learning_rate": 9.76995399079816e-06, "loss": 0.4791, "step": 582 }, { "epoch": 0.18, "learning_rate": 9.769553910782156e-06, "loss": 0.2225, "step": 583 }, { "epoch": 0.18, "learning_rate": 9.769153830766153e-06, "loss": 0.6271, "step": 584 }, { "epoch": 0.18, "learning_rate": 9.76875375075015e-06, "loss": 0.386, "step": 585 }, { "epoch": 0.18, "learning_rate": 9.768353670734148e-06, "loss": 0.4709, "step": 586 }, { "epoch": 0.18, "learning_rate": 9.767953590718145e-06, "loss": 0.5064, "step": 587 }, { "epoch": 0.18, "learning_rate": 9.767553510702142e-06, "loss": 0.3059, "step": 588 }, { "epoch": 0.18, "learning_rate": 9.767153430686138e-06, "loss": 0.3185, "step": 589 }, { "epoch": 0.18, "learning_rate": 9.766753350670135e-06, "loss": 0.4981, "step": 590 }, { "epoch": 0.18, "learning_rate": 9.766353270654132e-06, "loss": 0.463, "step": 591 }, { "epoch": 0.18, "learning_rate": 9.765953190638128e-06, "loss": 0.4845, "step": 592 }, { "epoch": 0.18, "learning_rate": 9.765553110622125e-06, "loss": 0.3817, "step": 593 }, { "epoch": 0.18, "learning_rate": 9.765153030606122e-06, "loss": 0.4445, "step": 594 }, { "epoch": 0.18, "learning_rate": 9.764752950590118e-06, "loss": 0.4736, "step": 595 }, { "epoch": 0.18, "learning_rate": 9.764352870574115e-06, "loss": 0.3502, "step": 596 }, { "epoch": 0.19, "learning_rate": 9.763952790558112e-06, "loss": 0.2861, "step": 597 }, { "epoch": 0.19, "learning_rate": 9.763552710542108e-06, "loss": 0.1695, "step": 598 }, { "epoch": 0.19, "learning_rate": 9.763152630526105e-06, "loss": 0.4166, "step": 599 }, { "epoch": 0.19, "learning_rate": 9.762752550510104e-06, "loss": 0.329, "step": 600 }, { "epoch": 0.19, "learning_rate": 9.7623524704941e-06, "loss": 0.4227, "step": 601 }, { "epoch": 0.19, "learning_rate": 9.761952390478097e-06, "loss": 0.2358, "step": 602 }, { "epoch": 0.19, "learning_rate": 9.761552310462094e-06, "loss": 0.3, "step": 603 }, { "epoch": 0.19, "learning_rate": 9.76115223044609e-06, "loss": 0.2194, "step": 604 }, { "epoch": 0.19, "learning_rate": 9.760752150430087e-06, "loss": 0.2054, "step": 605 }, { "epoch": 0.19, "learning_rate": 9.760352070414084e-06, "loss": 0.3609, "step": 606 }, { "epoch": 0.19, "learning_rate": 9.75995199039808e-06, "loss": 0.3409, "step": 607 }, { "epoch": 0.19, "learning_rate": 9.759551910382077e-06, "loss": 0.2268, "step": 608 }, { "epoch": 0.19, "learning_rate": 9.759151830366074e-06, "loss": 0.4185, "step": 609 }, { "epoch": 0.19, "learning_rate": 9.75875175035007e-06, "loss": 0.2515, "step": 610 }, { "epoch": 0.19, "learning_rate": 9.758351670334067e-06, "loss": 0.4358, "step": 611 }, { "epoch": 0.19, "learning_rate": 9.757951590318064e-06, "loss": 0.4391, "step": 612 }, { "epoch": 0.19, "learning_rate": 9.75755151030206e-06, "loss": 0.3822, "step": 613 }, { "epoch": 0.19, "learning_rate": 9.757151430286059e-06, "loss": 0.3096, "step": 614 }, { "epoch": 0.19, "learning_rate": 9.756751350270056e-06, "loss": 0.3551, "step": 615 }, { "epoch": 0.19, "learning_rate": 9.756351270254052e-06, "loss": 0.5754, "step": 616 }, { "epoch": 0.19, "learning_rate": 9.755951190238049e-06, "loss": 0.2865, "step": 617 }, { "epoch": 0.19, "learning_rate": 9.755551110222046e-06, "loss": 0.3589, "step": 618 }, { "epoch": 0.19, "learning_rate": 9.755151030206042e-06, "loss": 0.2692, "step": 619 }, { "epoch": 0.19, "learning_rate": 9.754750950190039e-06, "loss": 0.3683, "step": 620 }, { "epoch": 0.19, "learning_rate": 9.754350870174036e-06, "loss": 0.452, "step": 621 }, { "epoch": 0.19, "learning_rate": 9.753950790158032e-06, "loss": 0.3681, "step": 622 }, { "epoch": 0.19, "learning_rate": 9.753550710142029e-06, "loss": 0.4322, "step": 623 }, { "epoch": 0.19, "learning_rate": 9.753150630126026e-06, "loss": 0.3847, "step": 624 }, { "epoch": 0.19, "learning_rate": 9.752750550110022e-06, "loss": 0.3555, "step": 625 }, { "epoch": 0.19, "learning_rate": 9.752350470094019e-06, "loss": 0.4345, "step": 626 }, { "epoch": 0.19, "learning_rate": 9.751950390078016e-06, "loss": 0.5827, "step": 627 }, { "epoch": 0.19, "learning_rate": 9.751550310062014e-06, "loss": 0.4008, "step": 628 }, { "epoch": 0.2, "learning_rate": 9.75115023004601e-06, "loss": 0.2943, "step": 629 }, { "epoch": 0.2, "learning_rate": 9.750750150030008e-06, "loss": 0.434, "step": 630 }, { "epoch": 0.2, "learning_rate": 9.750350070014004e-06, "loss": 0.2285, "step": 631 }, { "epoch": 0.2, "learning_rate": 9.749949989998e-06, "loss": 0.2346, "step": 632 }, { "epoch": 0.2, "learning_rate": 9.749549909981998e-06, "loss": 0.3656, "step": 633 }, { "epoch": 0.2, "learning_rate": 9.749149829965994e-06, "loss": 0.4561, "step": 634 }, { "epoch": 0.2, "learning_rate": 9.748749749949991e-06, "loss": 0.3175, "step": 635 }, { "epoch": 0.2, "learning_rate": 9.748349669933988e-06, "loss": 0.41, "step": 636 }, { "epoch": 0.2, "learning_rate": 9.747949589917984e-06, "loss": 0.3452, "step": 637 }, { "epoch": 0.2, "learning_rate": 9.747549509901981e-06, "loss": 0.3973, "step": 638 }, { "epoch": 0.2, "learning_rate": 9.747149429885978e-06, "loss": 0.3089, "step": 639 }, { "epoch": 0.2, "learning_rate": 9.746749349869974e-06, "loss": 0.4479, "step": 640 }, { "epoch": 0.2, "learning_rate": 9.746349269853971e-06, "loss": 0.6035, "step": 641 }, { "epoch": 0.2, "learning_rate": 9.74594918983797e-06, "loss": 0.336, "step": 642 }, { "epoch": 0.2, "learning_rate": 9.745549109821966e-06, "loss": 0.2299, "step": 643 }, { "epoch": 0.2, "learning_rate": 9.745149029805961e-06, "loss": 0.3514, "step": 644 }, { "epoch": 0.2, "learning_rate": 9.744748949789958e-06, "loss": 0.3408, "step": 645 }, { "epoch": 0.2, "learning_rate": 9.744348869773954e-06, "loss": 0.3488, "step": 646 }, { "epoch": 0.2, "learning_rate": 9.743948789757953e-06, "loss": 0.3735, "step": 647 }, { "epoch": 0.2, "learning_rate": 9.74354870974195e-06, "loss": 0.3445, "step": 648 }, { "epoch": 0.2, "learning_rate": 9.743148629725946e-06, "loss": 0.3156, "step": 649 }, { "epoch": 0.2, "learning_rate": 9.742748549709943e-06, "loss": 0.5296, "step": 650 }, { "epoch": 0.2, "learning_rate": 9.74234846969394e-06, "loss": 0.3592, "step": 651 }, { "epoch": 0.2, "learning_rate": 9.741948389677936e-06, "loss": 0.2448, "step": 652 }, { "epoch": 0.2, "learning_rate": 9.741548309661933e-06, "loss": 0.4387, "step": 653 }, { "epoch": 0.2, "learning_rate": 9.74114822964593e-06, "loss": 0.4539, "step": 654 }, { "epoch": 0.2, "learning_rate": 9.740748149629926e-06, "loss": 0.5028, "step": 655 }, { "epoch": 0.2, "learning_rate": 9.740348069613925e-06, "loss": 0.3034, "step": 656 }, { "epoch": 0.2, "learning_rate": 9.73994798959792e-06, "loss": 0.3181, "step": 657 }, { "epoch": 0.2, "learning_rate": 9.739547909581916e-06, "loss": 0.5011, "step": 658 }, { "epoch": 0.2, "learning_rate": 9.739147829565913e-06, "loss": 0.4156, "step": 659 }, { "epoch": 0.2, "learning_rate": 9.73874774954991e-06, "loss": 0.1809, "step": 660 }, { "epoch": 0.21, "learning_rate": 9.738347669533908e-06, "loss": 0.4385, "step": 661 }, { "epoch": 0.21, "learning_rate": 9.737947589517905e-06, "loss": 0.2085, "step": 662 }, { "epoch": 0.21, "learning_rate": 9.737547509501902e-06, "loss": 0.4171, "step": 663 }, { "epoch": 0.21, "learning_rate": 9.737147429485898e-06, "loss": 0.2214, "step": 664 }, { "epoch": 0.21, "learning_rate": 9.736747349469895e-06, "loss": 0.2262, "step": 665 }, { "epoch": 0.21, "learning_rate": 9.736347269453892e-06, "loss": 0.342, "step": 666 }, { "epoch": 0.21, "learning_rate": 9.735947189437888e-06, "loss": 0.4043, "step": 667 }, { "epoch": 0.21, "learning_rate": 9.735547109421885e-06, "loss": 0.2634, "step": 668 }, { "epoch": 0.21, "learning_rate": 9.735147029405883e-06, "loss": 0.4191, "step": 669 }, { "epoch": 0.21, "learning_rate": 9.734746949389878e-06, "loss": 0.5036, "step": 670 }, { "epoch": 0.21, "learning_rate": 9.734346869373875e-06, "loss": 0.3686, "step": 671 }, { "epoch": 0.21, "learning_rate": 9.733946789357872e-06, "loss": 0.3531, "step": 672 }, { "epoch": 0.21, "learning_rate": 9.733546709341868e-06, "loss": 0.3459, "step": 673 }, { "epoch": 0.21, "learning_rate": 9.733146629325865e-06, "loss": 0.3802, "step": 674 }, { "epoch": 0.21, "learning_rate": 9.732746549309863e-06, "loss": 0.4027, "step": 675 }, { "epoch": 0.21, "learning_rate": 9.73234646929386e-06, "loss": 0.3219, "step": 676 }, { "epoch": 0.21, "learning_rate": 9.731946389277857e-06, "loss": 0.4532, "step": 677 }, { "epoch": 0.21, "learning_rate": 9.731546309261854e-06, "loss": 0.2836, "step": 678 }, { "epoch": 0.21, "learning_rate": 9.73114622924585e-06, "loss": 0.5471, "step": 679 }, { "epoch": 0.21, "learning_rate": 9.730746149229847e-06, "loss": 0.2268, "step": 680 }, { "epoch": 0.21, "learning_rate": 9.730346069213844e-06, "loss": 0.2307, "step": 681 }, { "epoch": 0.21, "learning_rate": 9.72994598919784e-06, "loss": 0.3851, "step": 682 }, { "epoch": 0.21, "learning_rate": 9.729545909181837e-06, "loss": 0.208, "step": 683 }, { "epoch": 0.21, "learning_rate": 9.729145829165834e-06, "loss": 0.3298, "step": 684 }, { "epoch": 0.21, "learning_rate": 9.72874574914983e-06, "loss": 0.2647, "step": 685 }, { "epoch": 0.21, "learning_rate": 9.728345669133827e-06, "loss": 0.2748, "step": 686 }, { "epoch": 0.21, "learning_rate": 9.727945589117824e-06, "loss": 0.3178, "step": 687 }, { "epoch": 0.21, "learning_rate": 9.72754550910182e-06, "loss": 0.3835, "step": 688 }, { "epoch": 0.21, "learning_rate": 9.727145429085819e-06, "loss": 0.3473, "step": 689 }, { "epoch": 0.21, "learning_rate": 9.726745349069815e-06, "loss": 0.281, "step": 690 }, { "epoch": 0.21, "learning_rate": 9.726345269053812e-06, "loss": 0.3063, "step": 691 }, { "epoch": 0.21, "learning_rate": 9.725945189037809e-06, "loss": 0.445, "step": 692 }, { "epoch": 0.22, "learning_rate": 9.725545109021806e-06, "loss": 0.3038, "step": 693 }, { "epoch": 0.22, "learning_rate": 9.725145029005802e-06, "loss": 0.3236, "step": 694 }, { "epoch": 0.22, "learning_rate": 9.724744948989799e-06, "loss": 0.5263, "step": 695 }, { "epoch": 0.22, "learning_rate": 9.724344868973796e-06, "loss": 0.4574, "step": 696 }, { "epoch": 0.22, "learning_rate": 9.723944788957792e-06, "loss": 0.3638, "step": 697 }, { "epoch": 0.22, "learning_rate": 9.723544708941789e-06, "loss": 0.273, "step": 698 }, { "epoch": 0.22, "learning_rate": 9.723144628925786e-06, "loss": 0.4073, "step": 699 }, { "epoch": 0.22, "learning_rate": 9.722744548909782e-06, "loss": 0.2995, "step": 700 }, { "epoch": 0.22, "learning_rate": 9.722344468893779e-06, "loss": 0.3238, "step": 701 }, { "epoch": 0.22, "learning_rate": 9.721944388877776e-06, "loss": 0.5982, "step": 702 }, { "epoch": 0.22, "learning_rate": 9.721544308861774e-06, "loss": 0.4448, "step": 703 }, { "epoch": 0.22, "learning_rate": 9.72114422884577e-06, "loss": 0.3163, "step": 704 }, { "epoch": 0.22, "learning_rate": 9.720744148829767e-06, "loss": 0.2644, "step": 705 }, { "epoch": 0.22, "learning_rate": 9.720344068813764e-06, "loss": 0.4278, "step": 706 }, { "epoch": 0.22, "learning_rate": 9.719943988797759e-06, "loss": 0.4367, "step": 707 }, { "epoch": 0.22, "learning_rate": 9.719543908781758e-06, "loss": 0.1879, "step": 708 }, { "epoch": 0.22, "learning_rate": 9.719143828765754e-06, "loss": 0.4286, "step": 709 }, { "epoch": 0.22, "learning_rate": 9.718743748749751e-06, "loss": 0.3507, "step": 710 }, { "epoch": 0.22, "learning_rate": 9.718343668733748e-06, "loss": 0.2779, "step": 711 }, { "epoch": 0.22, "learning_rate": 9.717943588717744e-06, "loss": 0.3484, "step": 712 }, { "epoch": 0.22, "learning_rate": 9.717543508701741e-06, "loss": 0.3698, "step": 713 }, { "epoch": 0.22, "learning_rate": 9.717143428685738e-06, "loss": 0.4307, "step": 714 }, { "epoch": 0.22, "learning_rate": 9.716743348669734e-06, "loss": 0.2927, "step": 715 }, { "epoch": 0.22, "learning_rate": 9.716343268653731e-06, "loss": 0.4766, "step": 716 }, { "epoch": 0.22, "learning_rate": 9.71594318863773e-06, "loss": 0.3482, "step": 717 }, { "epoch": 0.22, "learning_rate": 9.715543108621726e-06, "loss": 0.3828, "step": 718 }, { "epoch": 0.22, "learning_rate": 9.715143028605723e-06, "loss": 0.3484, "step": 719 }, { "epoch": 0.22, "learning_rate": 9.714742948589718e-06, "loss": 0.247, "step": 720 }, { "epoch": 0.22, "learning_rate": 9.714342868573714e-06, "loss": 0.317, "step": 721 }, { "epoch": 0.22, "learning_rate": 9.713942788557713e-06, "loss": 0.3965, "step": 722 }, { "epoch": 0.22, "learning_rate": 9.71354270854171e-06, "loss": 0.5274, "step": 723 }, { "epoch": 0.22, "learning_rate": 9.713142628525706e-06, "loss": 0.4602, "step": 724 }, { "epoch": 0.23, "learning_rate": 9.712742548509703e-06, "loss": 0.4132, "step": 725 }, { "epoch": 0.23, "learning_rate": 9.7123424684937e-06, "loss": 0.4189, "step": 726 }, { "epoch": 0.23, "learning_rate": 9.711942388477696e-06, "loss": 0.3526, "step": 727 }, { "epoch": 0.23, "learning_rate": 9.711542308461693e-06, "loss": 0.3149, "step": 728 }, { "epoch": 0.23, "learning_rate": 9.71114222844569e-06, "loss": 0.3723, "step": 729 }, { "epoch": 0.23, "learning_rate": 9.710742148429686e-06, "loss": 0.2849, "step": 730 }, { "epoch": 0.23, "learning_rate": 9.710342068413685e-06, "loss": 0.2896, "step": 731 }, { "epoch": 0.23, "learning_rate": 9.709941988397681e-06, "loss": 0.5768, "step": 732 }, { "epoch": 0.23, "learning_rate": 9.709541908381676e-06, "loss": 0.4367, "step": 733 }, { "epoch": 0.23, "learning_rate": 9.709141828365673e-06, "loss": 0.2134, "step": 734 }, { "epoch": 0.23, "learning_rate": 9.70874174834967e-06, "loss": 0.3658, "step": 735 }, { "epoch": 0.23, "learning_rate": 9.708341668333668e-06, "loss": 0.1966, "step": 736 }, { "epoch": 0.23, "learning_rate": 9.707941588317665e-06, "loss": 0.4411, "step": 737 }, { "epoch": 0.23, "learning_rate": 9.707541508301661e-06, "loss": 0.3301, "step": 738 }, { "epoch": 0.23, "learning_rate": 9.707141428285658e-06, "loss": 0.4683, "step": 739 }, { "epoch": 0.23, "learning_rate": 9.706741348269655e-06, "loss": 0.2812, "step": 740 }, { "epoch": 0.23, "learning_rate": 9.706341268253652e-06, "loss": 0.2616, "step": 741 }, { "epoch": 0.23, "learning_rate": 9.705941188237648e-06, "loss": 0.4347, "step": 742 }, { "epoch": 0.23, "learning_rate": 9.705541108221645e-06, "loss": 0.373, "step": 743 }, { "epoch": 0.23, "learning_rate": 9.705141028205642e-06, "loss": 0.321, "step": 744 }, { "epoch": 0.23, "learning_rate": 9.704740948189638e-06, "loss": 0.3364, "step": 745 }, { "epoch": 0.23, "learning_rate": 9.704340868173635e-06, "loss": 0.3612, "step": 746 }, { "epoch": 0.23, "learning_rate": 9.703940788157632e-06, "loss": 0.2907, "step": 747 }, { "epoch": 0.23, "learning_rate": 9.703540708141628e-06, "loss": 0.2709, "step": 748 }, { "epoch": 0.23, "learning_rate": 9.703140628125625e-06, "loss": 0.39, "step": 749 }, { "epoch": 0.23, "learning_rate": 9.702740548109623e-06, "loss": 0.3944, "step": 750 }, { "epoch": 0.23, "learning_rate": 9.70234046809362e-06, "loss": 0.2553, "step": 751 }, { "epoch": 0.23, "learning_rate": 9.701940388077617e-06, "loss": 0.3964, "step": 752 }, { "epoch": 0.23, "learning_rate": 9.701540308061613e-06, "loss": 0.3441, "step": 753 }, { "epoch": 0.23, "learning_rate": 9.70114022804561e-06, "loss": 0.3353, "step": 754 }, { "epoch": 0.23, "learning_rate": 9.700740148029607e-06, "loss": 0.2981, "step": 755 }, { "epoch": 0.23, "learning_rate": 9.700340068013604e-06, "loss": 0.3797, "step": 756 }, { "epoch": 0.23, "learning_rate": 9.6999399879976e-06, "loss": 0.3403, "step": 757 }, { "epoch": 0.24, "learning_rate": 9.699539907981597e-06, "loss": 0.435, "step": 758 }, { "epoch": 0.24, "learning_rate": 9.699139827965594e-06, "loss": 0.2127, "step": 759 }, { "epoch": 0.24, "learning_rate": 9.69873974794959e-06, "loss": 0.266, "step": 760 }, { "epoch": 0.24, "learning_rate": 9.698339667933587e-06, "loss": 0.3455, "step": 761 }, { "epoch": 0.24, "learning_rate": 9.697939587917584e-06, "loss": 0.2911, "step": 762 }, { "epoch": 0.24, "learning_rate": 9.69753950790158e-06, "loss": 0.3778, "step": 763 }, { "epoch": 0.24, "learning_rate": 9.697139427885579e-06, "loss": 0.2823, "step": 764 }, { "epoch": 0.24, "learning_rate": 9.696739347869575e-06, "loss": 0.4198, "step": 765 }, { "epoch": 0.24, "learning_rate": 9.696339267853572e-06, "loss": 0.3512, "step": 766 }, { "epoch": 0.24, "learning_rate": 9.695939187837569e-06, "loss": 0.2047, "step": 767 }, { "epoch": 0.24, "learning_rate": 9.695539107821565e-06, "loss": 0.3418, "step": 768 }, { "epoch": 0.24, "learning_rate": 9.695139027805562e-06, "loss": 0.4446, "step": 769 }, { "epoch": 0.24, "learning_rate": 9.694738947789559e-06, "loss": 0.551, "step": 770 }, { "epoch": 0.24, "learning_rate": 9.694338867773556e-06, "loss": 0.4669, "step": 771 }, { "epoch": 0.24, "learning_rate": 9.693938787757552e-06, "loss": 0.3361, "step": 772 }, { "epoch": 0.24, "learning_rate": 9.693538707741549e-06, "loss": 0.4452, "step": 773 }, { "epoch": 0.24, "learning_rate": 9.693138627725546e-06, "loss": 0.4813, "step": 774 }, { "epoch": 0.24, "learning_rate": 9.692738547709542e-06, "loss": 0.4461, "step": 775 }, { "epoch": 0.24, "learning_rate": 9.692338467693539e-06, "loss": 0.2942, "step": 776 }, { "epoch": 0.24, "learning_rate": 9.691938387677536e-06, "loss": 0.27, "step": 777 }, { "epoch": 0.24, "learning_rate": 9.691538307661534e-06, "loss": 0.375, "step": 778 }, { "epoch": 0.24, "learning_rate": 9.69113822764553e-06, "loss": 0.4068, "step": 779 }, { "epoch": 0.24, "learning_rate": 9.690738147629527e-06, "loss": 0.3507, "step": 780 }, { "epoch": 0.24, "learning_rate": 9.690338067613524e-06, "loss": 0.2737, "step": 781 }, { "epoch": 0.24, "learning_rate": 9.68993798759752e-06, "loss": 0.319, "step": 782 }, { "epoch": 0.24, "learning_rate": 9.689537907581516e-06, "loss": 0.2184, "step": 783 }, { "epoch": 0.24, "learning_rate": 9.689137827565514e-06, "loss": 0.3947, "step": 784 }, { "epoch": 0.24, "learning_rate": 9.68873774754951e-06, "loss": 0.3283, "step": 785 }, { "epoch": 0.24, "learning_rate": 9.688337667533508e-06, "loss": 0.211, "step": 786 }, { "epoch": 0.24, "learning_rate": 9.687937587517504e-06, "loss": 0.3276, "step": 787 }, { "epoch": 0.24, "learning_rate": 9.687537507501501e-06, "loss": 0.3517, "step": 788 }, { "epoch": 0.24, "learning_rate": 9.687137427485498e-06, "loss": 0.362, "step": 789 }, { "epoch": 0.25, "learning_rate": 9.686737347469494e-06, "loss": 0.43, "step": 790 }, { "epoch": 0.25, "learning_rate": 9.686337267453491e-06, "loss": 0.4082, "step": 791 }, { "epoch": 0.25, "learning_rate": 9.68593718743749e-06, "loss": 0.398, "step": 792 }, { "epoch": 0.25, "learning_rate": 9.685537107421486e-06, "loss": 0.3726, "step": 793 }, { "epoch": 0.25, "learning_rate": 9.685137027405483e-06, "loss": 0.3617, "step": 794 }, { "epoch": 0.25, "learning_rate": 9.68473694738948e-06, "loss": 0.4241, "step": 795 }, { "epoch": 0.25, "learning_rate": 9.684336867373474e-06, "loss": 0.4311, "step": 796 }, { "epoch": 0.25, "learning_rate": 9.683936787357471e-06, "loss": 0.3363, "step": 797 }, { "epoch": 0.25, "learning_rate": 9.68353670734147e-06, "loss": 0.4111, "step": 798 }, { "epoch": 0.25, "learning_rate": 9.683136627325466e-06, "loss": 0.3405, "step": 799 }, { "epoch": 0.25, "learning_rate": 9.682736547309463e-06, "loss": 0.3302, "step": 800 }, { "epoch": 0.25, "learning_rate": 9.68233646729346e-06, "loss": 0.5335, "step": 801 }, { "epoch": 0.25, "learning_rate": 9.681936387277456e-06, "loss": 0.6628, "step": 802 }, { "epoch": 0.25, "learning_rate": 9.681536307261453e-06, "loss": 0.4782, "step": 803 }, { "epoch": 0.25, "learning_rate": 9.68113622724545e-06, "loss": 0.3331, "step": 804 }, { "epoch": 0.25, "learning_rate": 9.680736147229446e-06, "loss": 0.6416, "step": 805 }, { "epoch": 0.25, "learning_rate": 9.680336067213445e-06, "loss": 0.2864, "step": 806 }, { "epoch": 0.25, "learning_rate": 9.679935987197441e-06, "loss": 0.4169, "step": 807 }, { "epoch": 0.25, "learning_rate": 9.679535907181436e-06, "loss": 0.3886, "step": 808 }, { "epoch": 0.25, "learning_rate": 9.679135827165433e-06, "loss": 0.3264, "step": 809 }, { "epoch": 0.25, "learning_rate": 9.67873574714943e-06, "loss": 0.361, "step": 810 }, { "epoch": 0.25, "learning_rate": 9.678335667133426e-06, "loss": 0.4322, "step": 811 }, { "epoch": 0.25, "learning_rate": 9.677935587117425e-06, "loss": 0.3764, "step": 812 }, { "epoch": 0.25, "learning_rate": 9.677535507101421e-06, "loss": 0.4813, "step": 813 }, { "epoch": 0.25, "learning_rate": 9.677135427085418e-06, "loss": 0.2033, "step": 814 }, { "epoch": 0.25, "learning_rate": 9.676735347069415e-06, "loss": 0.2755, "step": 815 }, { "epoch": 0.25, "learning_rate": 9.676335267053411e-06, "loss": 0.2222, "step": 816 }, { "epoch": 0.25, "learning_rate": 9.675935187037408e-06, "loss": 0.2729, "step": 817 }, { "epoch": 0.25, "learning_rate": 9.675535107021405e-06, "loss": 0.4077, "step": 818 }, { "epoch": 0.25, "learning_rate": 9.675135027005402e-06, "loss": 0.33, "step": 819 }, { "epoch": 0.25, "learning_rate": 9.6747349469894e-06, "loss": 0.3948, "step": 820 }, { "epoch": 0.25, "learning_rate": 9.674334866973395e-06, "loss": 0.3599, "step": 821 }, { "epoch": 0.26, "learning_rate": 9.673934786957392e-06, "loss": 0.2937, "step": 822 }, { "epoch": 0.26, "learning_rate": 9.673534706941388e-06, "loss": 0.4517, "step": 823 }, { "epoch": 0.26, "learning_rate": 9.673134626925385e-06, "loss": 0.2369, "step": 824 }, { "epoch": 0.26, "learning_rate": 9.672734546909382e-06, "loss": 0.2261, "step": 825 }, { "epoch": 0.26, "learning_rate": 9.67233446689338e-06, "loss": 0.422, "step": 826 }, { "epoch": 0.26, "learning_rate": 9.671934386877377e-06, "loss": 0.3529, "step": 827 }, { "epoch": 0.26, "learning_rate": 9.671534306861373e-06, "loss": 0.4301, "step": 828 }, { "epoch": 0.26, "learning_rate": 9.67113422684537e-06, "loss": 0.3549, "step": 829 }, { "epoch": 0.26, "learning_rate": 9.670734146829367e-06, "loss": 0.3666, "step": 830 }, { "epoch": 0.26, "learning_rate": 9.670334066813363e-06, "loss": 0.4036, "step": 831 }, { "epoch": 0.26, "learning_rate": 9.66993398679736e-06, "loss": 0.3029, "step": 832 }, { "epoch": 0.26, "learning_rate": 9.669533906781357e-06, "loss": 0.3063, "step": 833 }, { "epoch": 0.26, "learning_rate": 9.669133826765354e-06, "loss": 0.6304, "step": 834 }, { "epoch": 0.26, "learning_rate": 9.66873374674935e-06, "loss": 0.3827, "step": 835 }, { "epoch": 0.26, "learning_rate": 9.668333666733347e-06, "loss": 0.3083, "step": 836 }, { "epoch": 0.26, "learning_rate": 9.667933586717344e-06, "loss": 0.3091, "step": 837 }, { "epoch": 0.26, "learning_rate": 9.66753350670134e-06, "loss": 0.3899, "step": 838 }, { "epoch": 0.26, "learning_rate": 9.667133426685339e-06, "loss": 0.3504, "step": 839 }, { "epoch": 0.26, "learning_rate": 9.666733346669335e-06, "loss": 0.3686, "step": 840 }, { "epoch": 0.26, "learning_rate": 9.666333266653332e-06, "loss": 0.5078, "step": 841 }, { "epoch": 0.26, "learning_rate": 9.665933186637329e-06, "loss": 0.2779, "step": 842 }, { "epoch": 0.26, "learning_rate": 9.665533106621325e-06, "loss": 0.404, "step": 843 }, { "epoch": 0.26, "learning_rate": 9.665133026605322e-06, "loss": 0.2986, "step": 844 }, { "epoch": 0.26, "learning_rate": 9.664732946589319e-06, "loss": 0.3369, "step": 845 }, { "epoch": 0.26, "learning_rate": 9.664332866573315e-06, "loss": 0.2381, "step": 846 }, { "epoch": 0.26, "learning_rate": 9.663932786557312e-06, "loss": 0.3767, "step": 847 }, { "epoch": 0.26, "learning_rate": 9.663532706541309e-06, "loss": 0.2887, "step": 848 }, { "epoch": 0.26, "learning_rate": 9.663132626525306e-06, "loss": 0.2829, "step": 849 }, { "epoch": 0.26, "learning_rate": 9.662732546509302e-06, "loss": 0.2594, "step": 850 }, { "epoch": 0.26, "learning_rate": 9.662332466493299e-06, "loss": 0.2624, "step": 851 }, { "epoch": 0.26, "learning_rate": 9.661932386477296e-06, "loss": 0.3938, "step": 852 }, { "epoch": 0.26, "learning_rate": 9.661532306461294e-06, "loss": 0.1763, "step": 853 }, { "epoch": 0.27, "learning_rate": 9.66113222644529e-06, "loss": 0.353, "step": 854 }, { "epoch": 0.27, "learning_rate": 9.660732146429287e-06, "loss": 0.37, "step": 855 }, { "epoch": 0.27, "learning_rate": 9.660332066413284e-06, "loss": 0.3051, "step": 856 }, { "epoch": 0.27, "learning_rate": 9.65993198639728e-06, "loss": 0.5069, "step": 857 }, { "epoch": 0.27, "learning_rate": 9.659531906381277e-06, "loss": 0.378, "step": 858 }, { "epoch": 0.27, "learning_rate": 9.659131826365274e-06, "loss": 0.3117, "step": 859 }, { "epoch": 0.27, "learning_rate": 9.65873174634927e-06, "loss": 0.3962, "step": 860 }, { "epoch": 0.27, "learning_rate": 9.658331666333267e-06, "loss": 0.4842, "step": 861 }, { "epoch": 0.27, "learning_rate": 9.657931586317264e-06, "loss": 0.3616, "step": 862 }, { "epoch": 0.27, "learning_rate": 9.65753150630126e-06, "loss": 0.371, "step": 863 }, { "epoch": 0.27, "learning_rate": 9.657131426285258e-06, "loss": 0.3246, "step": 864 }, { "epoch": 0.27, "learning_rate": 9.656731346269254e-06, "loss": 0.3286, "step": 865 }, { "epoch": 0.27, "learning_rate": 9.656331266253251e-06, "loss": 0.3689, "step": 866 }, { "epoch": 0.27, "learning_rate": 9.65593118623725e-06, "loss": 0.3274, "step": 867 }, { "epoch": 0.27, "learning_rate": 9.655531106221246e-06, "loss": 0.3614, "step": 868 }, { "epoch": 0.27, "learning_rate": 9.655131026205243e-06, "loss": 0.4377, "step": 869 }, { "epoch": 0.27, "learning_rate": 9.65473094618924e-06, "loss": 0.3107, "step": 870 }, { "epoch": 0.27, "learning_rate": 9.654330866173234e-06, "loss": 0.404, "step": 871 }, { "epoch": 0.27, "learning_rate": 9.653930786157231e-06, "loss": 0.4142, "step": 872 }, { "epoch": 0.27, "learning_rate": 9.65353070614123e-06, "loss": 0.4246, "step": 873 }, { "epoch": 0.27, "learning_rate": 9.653130626125226e-06, "loss": 0.4211, "step": 874 }, { "epoch": 0.27, "learning_rate": 9.652730546109223e-06, "loss": 0.3768, "step": 875 }, { "epoch": 0.27, "learning_rate": 9.65233046609322e-06, "loss": 0.3376, "step": 876 }, { "epoch": 0.27, "learning_rate": 9.651930386077216e-06, "loss": 0.278, "step": 877 }, { "epoch": 0.27, "learning_rate": 9.651530306061213e-06, "loss": 0.313, "step": 878 }, { "epoch": 0.27, "learning_rate": 9.65113022604521e-06, "loss": 0.4114, "step": 879 }, { "epoch": 0.27, "learning_rate": 9.650730146029206e-06, "loss": 0.3158, "step": 880 }, { "epoch": 0.27, "learning_rate": 9.650330066013205e-06, "loss": 0.2469, "step": 881 }, { "epoch": 0.27, "learning_rate": 9.649929985997201e-06, "loss": 0.3629, "step": 882 }, { "epoch": 0.27, "learning_rate": 9.649529905981198e-06, "loss": 0.2725, "step": 883 }, { "epoch": 0.27, "learning_rate": 9.649129825965193e-06, "loss": 0.298, "step": 884 }, { "epoch": 0.27, "learning_rate": 9.64872974594919e-06, "loss": 0.341, "step": 885 }, { "epoch": 0.27, "learning_rate": 9.648329665933186e-06, "loss": 0.3357, "step": 886 }, { "epoch": 0.28, "learning_rate": 9.647929585917185e-06, "loss": 0.4128, "step": 887 }, { "epoch": 0.28, "learning_rate": 9.647529505901181e-06, "loss": 0.1897, "step": 888 }, { "epoch": 0.28, "learning_rate": 9.647129425885178e-06, "loss": 0.3095, "step": 889 }, { "epoch": 0.28, "learning_rate": 9.646729345869175e-06, "loss": 0.4232, "step": 890 }, { "epoch": 0.28, "learning_rate": 9.646329265853171e-06, "loss": 0.349, "step": 891 }, { "epoch": 0.28, "learning_rate": 9.645929185837168e-06, "loss": 0.3167, "step": 892 }, { "epoch": 0.28, "learning_rate": 9.645529105821165e-06, "loss": 0.3356, "step": 893 }, { "epoch": 0.28, "learning_rate": 9.645129025805161e-06, "loss": 0.2963, "step": 894 }, { "epoch": 0.28, "learning_rate": 9.64472894578916e-06, "loss": 0.6336, "step": 895 }, { "epoch": 0.28, "learning_rate": 9.644328865773157e-06, "loss": 0.324, "step": 896 }, { "epoch": 0.28, "learning_rate": 9.643928785757152e-06, "loss": 0.2447, "step": 897 }, { "epoch": 0.28, "learning_rate": 9.643528705741148e-06, "loss": 0.2334, "step": 898 }, { "epoch": 0.28, "learning_rate": 9.643128625725145e-06, "loss": 0.3336, "step": 899 }, { "epoch": 0.28, "learning_rate": 9.642728545709142e-06, "loss": 0.4418, "step": 900 }, { "epoch": 0.28, "learning_rate": 9.64232846569314e-06, "loss": 0.4325, "step": 901 }, { "epoch": 0.28, "learning_rate": 9.641928385677137e-06, "loss": 0.3786, "step": 902 }, { "epoch": 0.28, "learning_rate": 9.641528305661133e-06, "loss": 0.448, "step": 903 }, { "epoch": 0.28, "learning_rate": 9.64112822564513e-06, "loss": 0.3707, "step": 904 }, { "epoch": 0.28, "learning_rate": 9.640728145629127e-06, "loss": 0.2606, "step": 905 }, { "epoch": 0.28, "learning_rate": 9.640328065613123e-06, "loss": 0.3756, "step": 906 }, { "epoch": 0.28, "learning_rate": 9.63992798559712e-06, "loss": 0.2758, "step": 907 }, { "epoch": 0.28, "learning_rate": 9.639527905581117e-06, "loss": 0.2843, "step": 908 }, { "epoch": 0.28, "learning_rate": 9.639127825565113e-06, "loss": 0.2912, "step": 909 }, { "epoch": 0.28, "learning_rate": 9.63872774554911e-06, "loss": 0.2373, "step": 910 }, { "epoch": 0.28, "learning_rate": 9.638327665533107e-06, "loss": 0.3225, "step": 911 }, { "epoch": 0.28, "learning_rate": 9.637927585517104e-06, "loss": 0.6533, "step": 912 }, { "epoch": 0.28, "learning_rate": 9.6375275055011e-06, "loss": 0.3464, "step": 913 }, { "epoch": 0.28, "learning_rate": 9.637127425485097e-06, "loss": 0.4854, "step": 914 }, { "epoch": 0.28, "learning_rate": 9.636727345469095e-06, "loss": 0.4858, "step": 915 }, { "epoch": 0.28, "learning_rate": 9.636327265453092e-06, "loss": 0.2283, "step": 916 }, { "epoch": 0.28, "learning_rate": 9.635927185437089e-06, "loss": 0.5253, "step": 917 }, { "epoch": 0.28, "learning_rate": 9.635527105421085e-06, "loss": 0.2808, "step": 918 }, { "epoch": 0.29, "learning_rate": 9.635127025405082e-06, "loss": 0.3531, "step": 919 }, { "epoch": 0.29, "learning_rate": 9.634726945389079e-06, "loss": 0.2158, "step": 920 }, { "epoch": 0.29, "learning_rate": 9.634326865373075e-06, "loss": 0.1837, "step": 921 }, { "epoch": 0.29, "learning_rate": 9.633926785357072e-06, "loss": 0.3951, "step": 922 }, { "epoch": 0.29, "learning_rate": 9.633526705341069e-06, "loss": 0.3628, "step": 923 }, { "epoch": 0.29, "learning_rate": 9.633126625325065e-06, "loss": 0.4418, "step": 924 }, { "epoch": 0.29, "learning_rate": 9.632726545309062e-06, "loss": 0.3231, "step": 925 }, { "epoch": 0.29, "learning_rate": 9.632326465293059e-06, "loss": 0.3343, "step": 926 }, { "epoch": 0.29, "learning_rate": 9.631926385277056e-06, "loss": 0.3215, "step": 927 }, { "epoch": 0.29, "learning_rate": 9.631526305261052e-06, "loss": 0.254, "step": 928 }, { "epoch": 0.29, "learning_rate": 9.63112622524505e-06, "loss": 0.2693, "step": 929 }, { "epoch": 0.29, "learning_rate": 9.630726145229047e-06, "loss": 0.3727, "step": 930 }, { "epoch": 0.29, "learning_rate": 9.630326065213044e-06, "loss": 0.4029, "step": 931 }, { "epoch": 0.29, "learning_rate": 9.62992598519704e-06, "loss": 0.3922, "step": 932 }, { "epoch": 0.29, "learning_rate": 9.629525905181037e-06, "loss": 0.2884, "step": 933 }, { "epoch": 0.29, "learning_rate": 9.629125825165034e-06, "loss": 0.2856, "step": 934 }, { "epoch": 0.29, "learning_rate": 9.62872574514903e-06, "loss": 0.3131, "step": 935 }, { "epoch": 0.29, "learning_rate": 9.628325665133027e-06, "loss": 0.3659, "step": 936 }, { "epoch": 0.29, "learning_rate": 9.627925585117024e-06, "loss": 0.4361, "step": 937 }, { "epoch": 0.29, "learning_rate": 9.62752550510102e-06, "loss": 0.3205, "step": 938 }, { "epoch": 0.29, "learning_rate": 9.627125425085017e-06, "loss": 0.3358, "step": 939 }, { "epoch": 0.29, "learning_rate": 9.626725345069014e-06, "loss": 0.245, "step": 940 }, { "epoch": 0.29, "learning_rate": 9.62632526505301e-06, "loss": 0.3203, "step": 941 }, { "epoch": 0.29, "learning_rate": 9.625925185037008e-06, "loss": 0.323, "step": 942 }, { "epoch": 0.29, "learning_rate": 9.625525105021006e-06, "loss": 0.2401, "step": 943 }, { "epoch": 0.29, "learning_rate": 9.625125025005003e-06, "loss": 0.3076, "step": 944 }, { "epoch": 0.29, "learning_rate": 9.624724944989e-06, "loss": 0.3324, "step": 945 }, { "epoch": 0.29, "learning_rate": 9.624324864972996e-06, "loss": 0.2104, "step": 946 }, { "epoch": 0.29, "learning_rate": 9.623924784956991e-06, "loss": 0.339, "step": 947 }, { "epoch": 0.29, "learning_rate": 9.62352470494099e-06, "loss": 0.2064, "step": 948 }, { "epoch": 0.29, "learning_rate": 9.623124624924986e-06, "loss": 0.5758, "step": 949 }, { "epoch": 0.29, "learning_rate": 9.622724544908983e-06, "loss": 0.2604, "step": 950 }, { "epoch": 0.3, "learning_rate": 9.62232446489298e-06, "loss": 0.348, "step": 951 }, { "epoch": 0.3, "learning_rate": 9.621924384876976e-06, "loss": 0.1638, "step": 952 }, { "epoch": 0.3, "learning_rate": 9.621524304860973e-06, "loss": 0.4152, "step": 953 }, { "epoch": 0.3, "learning_rate": 9.62112422484497e-06, "loss": 0.4569, "step": 954 }, { "epoch": 0.3, "learning_rate": 9.620724144828966e-06, "loss": 0.352, "step": 955 }, { "epoch": 0.3, "learning_rate": 9.620324064812963e-06, "loss": 0.1656, "step": 956 }, { "epoch": 0.3, "learning_rate": 9.619923984796961e-06, "loss": 0.228, "step": 957 }, { "epoch": 0.3, "learning_rate": 9.619523904780958e-06, "loss": 0.3713, "step": 958 }, { "epoch": 0.3, "learning_rate": 9.619123824764955e-06, "loss": 0.3236, "step": 959 }, { "epoch": 0.3, "learning_rate": 9.61872374474895e-06, "loss": 0.3355, "step": 960 }, { "epoch": 0.3, "learning_rate": 9.618323664732946e-06, "loss": 0.2398, "step": 961 }, { "epoch": 0.3, "learning_rate": 9.617923584716945e-06, "loss": 0.3105, "step": 962 }, { "epoch": 0.3, "learning_rate": 9.617523504700941e-06, "loss": 0.4283, "step": 963 }, { "epoch": 0.3, "learning_rate": 9.617123424684938e-06, "loss": 0.3844, "step": 964 }, { "epoch": 0.3, "learning_rate": 9.616723344668935e-06, "loss": 0.4189, "step": 965 }, { "epoch": 0.3, "learning_rate": 9.616323264652931e-06, "loss": 0.3531, "step": 966 }, { "epoch": 0.3, "learning_rate": 9.615923184636928e-06, "loss": 0.2719, "step": 967 }, { "epoch": 0.3, "learning_rate": 9.615523104620925e-06, "loss": 0.3335, "step": 968 }, { "epoch": 0.3, "learning_rate": 9.615123024604921e-06, "loss": 0.4357, "step": 969 }, { "epoch": 0.3, "learning_rate": 9.61472294458892e-06, "loss": 0.441, "step": 970 }, { "epoch": 0.3, "learning_rate": 9.614322864572917e-06, "loss": 0.2868, "step": 971 }, { "epoch": 0.3, "learning_rate": 9.613922784556911e-06, "loss": 0.3074, "step": 972 }, { "epoch": 0.3, "learning_rate": 9.613522704540908e-06, "loss": 0.205, "step": 973 }, { "epoch": 0.3, "learning_rate": 9.613122624524905e-06, "loss": 0.2805, "step": 974 }, { "epoch": 0.3, "learning_rate": 9.612722544508902e-06, "loss": 0.349, "step": 975 }, { "epoch": 0.3, "learning_rate": 9.6123224644929e-06, "loss": 0.2357, "step": 976 }, { "epoch": 0.3, "learning_rate": 9.611922384476897e-06, "loss": 0.471, "step": 977 }, { "epoch": 0.3, "learning_rate": 9.611522304460893e-06, "loss": 0.4577, "step": 978 }, { "epoch": 0.3, "learning_rate": 9.61112222444489e-06, "loss": 0.2575, "step": 979 }, { "epoch": 0.3, "learning_rate": 9.610722144428887e-06, "loss": 0.341, "step": 980 }, { "epoch": 0.3, "learning_rate": 9.610322064412883e-06, "loss": 0.3519, "step": 981 }, { "epoch": 0.3, "learning_rate": 9.60992198439688e-06, "loss": 0.2812, "step": 982 }, { "epoch": 0.31, "learning_rate": 9.609521904380877e-06, "loss": 0.3196, "step": 983 }, { "epoch": 0.31, "learning_rate": 9.609121824364875e-06, "loss": 0.2714, "step": 984 }, { "epoch": 0.31, "learning_rate": 9.60872174434887e-06, "loss": 0.2896, "step": 985 }, { "epoch": 0.31, "learning_rate": 9.608321664332867e-06, "loss": 0.3874, "step": 986 }, { "epoch": 0.31, "learning_rate": 9.607921584316863e-06, "loss": 0.4442, "step": 987 }, { "epoch": 0.31, "learning_rate": 9.60752150430086e-06, "loss": 0.2327, "step": 988 }, { "epoch": 0.31, "learning_rate": 9.607121424284857e-06, "loss": 0.5069, "step": 989 }, { "epoch": 0.31, "learning_rate": 9.606721344268855e-06, "loss": 0.4083, "step": 990 }, { "epoch": 0.31, "learning_rate": 9.606321264252852e-06, "loss": 0.2832, "step": 991 }, { "epoch": 0.31, "learning_rate": 9.605921184236849e-06, "loss": 0.3446, "step": 992 }, { "epoch": 0.31, "learning_rate": 9.605521104220845e-06, "loss": 0.2692, "step": 993 }, { "epoch": 0.31, "learning_rate": 9.605121024204842e-06, "loss": 0.3301, "step": 994 }, { "epoch": 0.31, "learning_rate": 9.604720944188839e-06, "loss": 0.2795, "step": 995 }, { "epoch": 0.31, "learning_rate": 9.604320864172835e-06, "loss": 0.3123, "step": 996 }, { "epoch": 0.31, "learning_rate": 9.603920784156832e-06, "loss": 0.3005, "step": 997 }, { "epoch": 0.31, "learning_rate": 9.603520704140829e-06, "loss": 0.3366, "step": 998 }, { "epoch": 0.31, "learning_rate": 9.603120624124825e-06, "loss": 0.3142, "step": 999 }, { "epoch": 0.31, "learning_rate": 9.602720544108822e-06, "loss": 0.6467, "step": 1000 }, { "epoch": 0.31, "learning_rate": 9.602320464092819e-06, "loss": 0.2561, "step": 1001 }, { "epoch": 0.31, "learning_rate": 9.601920384076815e-06, "loss": 0.1753, "step": 1002 }, { "epoch": 0.31, "learning_rate": 9.601520304060812e-06, "loss": 0.3459, "step": 1003 }, { "epoch": 0.31, "learning_rate": 9.60112022404481e-06, "loss": 0.4146, "step": 1004 }, { "epoch": 0.31, "learning_rate": 9.600720144028807e-06, "loss": 0.3614, "step": 1005 }, { "epoch": 0.31, "learning_rate": 9.600320064012804e-06, "loss": 0.3608, "step": 1006 }, { "epoch": 0.31, "learning_rate": 9.5999199839968e-06, "loss": 0.2939, "step": 1007 }, { "epoch": 0.31, "learning_rate": 9.599519903980797e-06, "loss": 0.3276, "step": 1008 }, { "epoch": 0.31, "learning_rate": 9.599119823964794e-06, "loss": 0.2512, "step": 1009 }, { "epoch": 0.31, "learning_rate": 9.59871974394879e-06, "loss": 0.3398, "step": 1010 }, { "epoch": 0.31, "learning_rate": 9.598319663932787e-06, "loss": 0.4991, "step": 1011 }, { "epoch": 0.31, "learning_rate": 9.597919583916784e-06, "loss": 0.2268, "step": 1012 }, { "epoch": 0.31, "learning_rate": 9.59751950390078e-06, "loss": 0.1874, "step": 1013 }, { "epoch": 0.31, "learning_rate": 9.597119423884777e-06, "loss": 0.282, "step": 1014 }, { "epoch": 0.32, "learning_rate": 9.596719343868774e-06, "loss": 0.3646, "step": 1015 }, { "epoch": 0.32, "learning_rate": 9.59631926385277e-06, "loss": 0.2887, "step": 1016 }, { "epoch": 0.32, "learning_rate": 9.595919183836767e-06, "loss": 0.5349, "step": 1017 }, { "epoch": 0.32, "learning_rate": 9.595519103820766e-06, "loss": 0.4017, "step": 1018 }, { "epoch": 0.32, "learning_rate": 9.595119023804763e-06, "loss": 0.4029, "step": 1019 }, { "epoch": 0.32, "learning_rate": 9.59471894378876e-06, "loss": 0.3962, "step": 1020 }, { "epoch": 0.32, "learning_rate": 9.594318863772756e-06, "loss": 0.4249, "step": 1021 }, { "epoch": 0.32, "learning_rate": 9.593918783756753e-06, "loss": 0.3132, "step": 1022 }, { "epoch": 0.32, "learning_rate": 9.59351870374075e-06, "loss": 0.3412, "step": 1023 }, { "epoch": 0.32, "learning_rate": 9.593118623724746e-06, "loss": 0.2372, "step": 1024 }, { "epoch": 0.32, "learning_rate": 9.592718543708743e-06, "loss": 0.319, "step": 1025 }, { "epoch": 0.32, "learning_rate": 9.59231846369274e-06, "loss": 0.3786, "step": 1026 }, { "epoch": 0.32, "learning_rate": 9.591918383676736e-06, "loss": 0.1758, "step": 1027 }, { "epoch": 0.32, "learning_rate": 9.591518303660733e-06, "loss": 0.3006, "step": 1028 }, { "epoch": 0.32, "learning_rate": 9.59111822364473e-06, "loss": 0.3128, "step": 1029 }, { "epoch": 0.32, "learning_rate": 9.590718143628726e-06, "loss": 0.2894, "step": 1030 }, { "epoch": 0.32, "learning_rate": 9.590318063612723e-06, "loss": 0.3441, "step": 1031 }, { "epoch": 0.32, "learning_rate": 9.589917983596721e-06, "loss": 0.2731, "step": 1032 }, { "epoch": 0.32, "learning_rate": 9.589517903580718e-06, "loss": 0.4102, "step": 1033 }, { "epoch": 0.32, "learning_rate": 9.589117823564715e-06, "loss": 0.4514, "step": 1034 }, { "epoch": 0.32, "learning_rate": 9.58871774354871e-06, "loss": 0.3616, "step": 1035 }, { "epoch": 0.32, "learning_rate": 9.588317663532706e-06, "loss": 0.4425, "step": 1036 }, { "epoch": 0.32, "learning_rate": 9.587917583516705e-06, "loss": 0.357, "step": 1037 }, { "epoch": 0.32, "learning_rate": 9.587517503500701e-06, "loss": 0.6361, "step": 1038 }, { "epoch": 0.32, "learning_rate": 9.587117423484698e-06, "loss": 0.3618, "step": 1039 }, { "epoch": 0.32, "learning_rate": 9.586717343468695e-06, "loss": 0.3988, "step": 1040 }, { "epoch": 0.32, "learning_rate": 9.586317263452691e-06, "loss": 0.4135, "step": 1041 }, { "epoch": 0.32, "learning_rate": 9.585917183436688e-06, "loss": 0.3874, "step": 1042 }, { "epoch": 0.32, "learning_rate": 9.585517103420685e-06, "loss": 0.3377, "step": 1043 }, { "epoch": 0.32, "learning_rate": 9.585117023404681e-06, "loss": 0.272, "step": 1044 }, { "epoch": 0.32, "learning_rate": 9.584716943388678e-06, "loss": 0.2492, "step": 1045 }, { "epoch": 0.32, "learning_rate": 9.584316863372676e-06, "loss": 0.4272, "step": 1046 }, { "epoch": 0.32, "learning_rate": 9.583916783356673e-06, "loss": 0.3003, "step": 1047 }, { "epoch": 0.33, "learning_rate": 9.583516703340668e-06, "loss": 0.3601, "step": 1048 }, { "epoch": 0.33, "learning_rate": 9.583116623324665e-06, "loss": 0.3409, "step": 1049 }, { "epoch": 0.33, "learning_rate": 9.582716543308661e-06, "loss": 0.1742, "step": 1050 }, { "epoch": 0.33, "learning_rate": 9.58231646329266e-06, "loss": 0.4423, "step": 1051 }, { "epoch": 0.33, "learning_rate": 9.581916383276657e-06, "loss": 0.2224, "step": 1052 }, { "epoch": 0.33, "learning_rate": 9.581516303260653e-06, "loss": 0.4188, "step": 1053 }, { "epoch": 0.33, "learning_rate": 9.58111622324465e-06, "loss": 0.1752, "step": 1054 }, { "epoch": 0.33, "learning_rate": 9.580716143228647e-06, "loss": 0.373, "step": 1055 }, { "epoch": 0.33, "learning_rate": 9.580316063212643e-06, "loss": 0.3117, "step": 1056 }, { "epoch": 0.33, "learning_rate": 9.57991598319664e-06, "loss": 0.1734, "step": 1057 }, { "epoch": 0.33, "learning_rate": 9.579515903180637e-06, "loss": 0.318, "step": 1058 }, { "epoch": 0.33, "learning_rate": 9.579115823164633e-06, "loss": 0.3073, "step": 1059 }, { "epoch": 0.33, "learning_rate": 9.578715743148632e-06, "loss": 0.2193, "step": 1060 }, { "epoch": 0.33, "learning_rate": 9.578315663132627e-06, "loss": 0.4839, "step": 1061 }, { "epoch": 0.33, "learning_rate": 9.577915583116623e-06, "loss": 0.2153, "step": 1062 }, { "epoch": 0.33, "learning_rate": 9.57751550310062e-06, "loss": 0.242, "step": 1063 }, { "epoch": 0.33, "learning_rate": 9.577115423084617e-06, "loss": 0.3137, "step": 1064 }, { "epoch": 0.33, "learning_rate": 9.576715343068615e-06, "loss": 0.2913, "step": 1065 }, { "epoch": 0.33, "learning_rate": 9.576315263052612e-06, "loss": 0.3086, "step": 1066 }, { "epoch": 0.33, "learning_rate": 9.575915183036609e-06, "loss": 0.1791, "step": 1067 }, { "epoch": 0.33, "learning_rate": 9.575515103020605e-06, "loss": 0.2143, "step": 1068 }, { "epoch": 0.33, "learning_rate": 9.575115023004602e-06, "loss": 0.4604, "step": 1069 }, { "epoch": 0.33, "learning_rate": 9.574714942988599e-06, "loss": 0.3624, "step": 1070 }, { "epoch": 0.33, "learning_rate": 9.574314862972595e-06, "loss": 0.2945, "step": 1071 }, { "epoch": 0.33, "learning_rate": 9.573914782956592e-06, "loss": 0.487, "step": 1072 }, { "epoch": 0.33, "learning_rate": 9.573514702940589e-06, "loss": 0.2649, "step": 1073 }, { "epoch": 0.33, "learning_rate": 9.573114622924585e-06, "loss": 0.2288, "step": 1074 }, { "epoch": 0.33, "learning_rate": 9.572714542908582e-06, "loss": 0.4406, "step": 1075 }, { "epoch": 0.33, "learning_rate": 9.572314462892579e-06, "loss": 0.4024, "step": 1076 }, { "epoch": 0.33, "learning_rate": 9.571914382876575e-06, "loss": 0.3954, "step": 1077 }, { "epoch": 0.33, "learning_rate": 9.571514302860572e-06, "loss": 0.2919, "step": 1078 }, { "epoch": 0.33, "learning_rate": 9.57111422284457e-06, "loss": 0.3271, "step": 1079 }, { "epoch": 0.34, "learning_rate": 9.570714142828567e-06, "loss": 0.2575, "step": 1080 }, { "epoch": 0.34, "learning_rate": 9.570314062812564e-06, "loss": 0.536, "step": 1081 }, { "epoch": 0.34, "learning_rate": 9.56991398279656e-06, "loss": 0.2778, "step": 1082 }, { "epoch": 0.34, "learning_rate": 9.569513902780557e-06, "loss": 0.4208, "step": 1083 }, { "epoch": 0.34, "learning_rate": 9.569113822764554e-06, "loss": 0.2282, "step": 1084 }, { "epoch": 0.34, "learning_rate": 9.56871374274855e-06, "loss": 0.2482, "step": 1085 }, { "epoch": 0.34, "learning_rate": 9.568313662732547e-06, "loss": 0.2746, "step": 1086 }, { "epoch": 0.34, "learning_rate": 9.567913582716544e-06, "loss": 0.203, "step": 1087 }, { "epoch": 0.34, "learning_rate": 9.56751350270054e-06, "loss": 0.2943, "step": 1088 }, { "epoch": 0.34, "learning_rate": 9.567113422684537e-06, "loss": 0.4034, "step": 1089 }, { "epoch": 0.34, "learning_rate": 9.566713342668534e-06, "loss": 0.3907, "step": 1090 }, { "epoch": 0.34, "learning_rate": 9.56631326265253e-06, "loss": 0.3433, "step": 1091 }, { "epoch": 0.34, "learning_rate": 9.565913182636527e-06, "loss": 0.3347, "step": 1092 }, { "epoch": 0.34, "learning_rate": 9.565513102620526e-06, "loss": 0.2362, "step": 1093 }, { "epoch": 0.34, "learning_rate": 9.565113022604522e-06, "loss": 0.4782, "step": 1094 }, { "epoch": 0.34, "learning_rate": 9.56471294258852e-06, "loss": 0.3833, "step": 1095 }, { "epoch": 0.34, "learning_rate": 9.564312862572516e-06, "loss": 0.3119, "step": 1096 }, { "epoch": 0.34, "learning_rate": 9.563912782556513e-06, "loss": 0.2546, "step": 1097 }, { "epoch": 0.34, "learning_rate": 9.563512702540508e-06, "loss": 0.1877, "step": 1098 }, { "epoch": 0.34, "learning_rate": 9.563112622524506e-06, "loss": 0.4173, "step": 1099 }, { "epoch": 0.34, "learning_rate": 9.562712542508503e-06, "loss": 0.2266, "step": 1100 }, { "epoch": 0.34, "learning_rate": 9.5623124624925e-06, "loss": 0.2266, "step": 1101 }, { "epoch": 0.34, "learning_rate": 9.561912382476496e-06, "loss": 0.3469, "step": 1102 }, { "epoch": 0.34, "learning_rate": 9.561512302460493e-06, "loss": 0.2221, "step": 1103 }, { "epoch": 0.34, "learning_rate": 9.56111222244449e-06, "loss": 0.2136, "step": 1104 }, { "epoch": 0.34, "learning_rate": 9.560712142428486e-06, "loss": 0.2023, "step": 1105 }, { "epoch": 0.34, "learning_rate": 9.560312062412483e-06, "loss": 0.3265, "step": 1106 }, { "epoch": 0.34, "learning_rate": 9.559911982396481e-06, "loss": 0.1697, "step": 1107 }, { "epoch": 0.34, "learning_rate": 9.559511902380478e-06, "loss": 0.3164, "step": 1108 }, { "epoch": 0.34, "learning_rate": 9.559111822364474e-06, "loss": 0.2938, "step": 1109 }, { "epoch": 0.34, "learning_rate": 9.558711742348471e-06, "loss": 0.4334, "step": 1110 }, { "epoch": 0.34, "learning_rate": 9.558311662332466e-06, "loss": 0.4237, "step": 1111 }, { "epoch": 0.35, "learning_rate": 9.557911582316463e-06, "loss": 0.4342, "step": 1112 }, { "epoch": 0.35, "learning_rate": 9.557511502300461e-06, "loss": 0.3799, "step": 1113 }, { "epoch": 0.35, "learning_rate": 9.557111422284458e-06, "loss": 0.4894, "step": 1114 }, { "epoch": 0.35, "learning_rate": 9.556711342268455e-06, "loss": 0.2086, "step": 1115 }, { "epoch": 0.35, "learning_rate": 9.556311262252451e-06, "loss": 0.3428, "step": 1116 }, { "epoch": 0.35, "learning_rate": 9.555911182236448e-06, "loss": 0.3378, "step": 1117 }, { "epoch": 0.35, "learning_rate": 9.555511102220445e-06, "loss": 0.3827, "step": 1118 }, { "epoch": 0.35, "learning_rate": 9.555111022204441e-06, "loss": 0.3031, "step": 1119 }, { "epoch": 0.35, "learning_rate": 9.554710942188438e-06, "loss": 0.206, "step": 1120 }, { "epoch": 0.35, "learning_rate": 9.554310862172436e-06, "loss": 0.3512, "step": 1121 }, { "epoch": 0.35, "learning_rate": 9.553910782156433e-06, "loss": 0.4862, "step": 1122 }, { "epoch": 0.35, "learning_rate": 9.55351070214043e-06, "loss": 0.291, "step": 1123 }, { "epoch": 0.35, "learning_rate": 9.553110622124425e-06, "loss": 0.2968, "step": 1124 }, { "epoch": 0.35, "learning_rate": 9.552710542108421e-06, "loss": 0.2667, "step": 1125 }, { "epoch": 0.35, "learning_rate": 9.552310462092418e-06, "loss": 0.2716, "step": 1126 }, { "epoch": 0.35, "learning_rate": 9.551910382076417e-06, "loss": 0.3073, "step": 1127 }, { "epoch": 0.35, "learning_rate": 9.551510302060413e-06, "loss": 0.3131, "step": 1128 }, { "epoch": 0.35, "learning_rate": 9.55111022204441e-06, "loss": 0.2181, "step": 1129 }, { "epoch": 0.35, "learning_rate": 9.550710142028407e-06, "loss": 0.2069, "step": 1130 }, { "epoch": 0.35, "learning_rate": 9.550310062012403e-06, "loss": 0.2939, "step": 1131 }, { "epoch": 0.35, "learning_rate": 9.5499099819964e-06, "loss": 0.272, "step": 1132 }, { "epoch": 0.35, "learning_rate": 9.549509901980397e-06, "loss": 0.383, "step": 1133 }, { "epoch": 0.35, "learning_rate": 9.549109821964393e-06, "loss": 0.2615, "step": 1134 }, { "epoch": 0.35, "learning_rate": 9.548709741948392e-06, "loss": 0.4153, "step": 1135 }, { "epoch": 0.35, "learning_rate": 9.548309661932387e-06, "loss": 0.1784, "step": 1136 }, { "epoch": 0.35, "learning_rate": 9.547909581916383e-06, "loss": 0.3996, "step": 1137 }, { "epoch": 0.35, "learning_rate": 9.54750950190038e-06, "loss": 0.2922, "step": 1138 }, { "epoch": 0.35, "learning_rate": 9.547109421884377e-06, "loss": 0.2507, "step": 1139 }, { "epoch": 0.35, "learning_rate": 9.546709341868375e-06, "loss": 0.3134, "step": 1140 }, { "epoch": 0.35, "learning_rate": 9.546309261852372e-06, "loss": 0.2661, "step": 1141 }, { "epoch": 0.35, "learning_rate": 9.545909181836368e-06, "loss": 0.2001, "step": 1142 }, { "epoch": 0.35, "learning_rate": 9.545509101820365e-06, "loss": 0.3771, "step": 1143 }, { "epoch": 0.36, "learning_rate": 9.545109021804362e-06, "loss": 0.4038, "step": 1144 }, { "epoch": 0.36, "learning_rate": 9.544708941788359e-06, "loss": 0.4245, "step": 1145 }, { "epoch": 0.36, "learning_rate": 9.544308861772355e-06, "loss": 0.342, "step": 1146 }, { "epoch": 0.36, "learning_rate": 9.543908781756352e-06, "loss": 0.2941, "step": 1147 }, { "epoch": 0.36, "learning_rate": 9.543508701740349e-06, "loss": 0.399, "step": 1148 }, { "epoch": 0.36, "learning_rate": 9.543108621724345e-06, "loss": 0.347, "step": 1149 }, { "epoch": 0.36, "learning_rate": 9.542708541708342e-06, "loss": 0.3316, "step": 1150 }, { "epoch": 0.36, "learning_rate": 9.542308461692339e-06, "loss": 0.3461, "step": 1151 }, { "epoch": 0.36, "learning_rate": 9.541908381676335e-06, "loss": 0.4296, "step": 1152 }, { "epoch": 0.36, "learning_rate": 9.541508301660332e-06, "loss": 0.2629, "step": 1153 }, { "epoch": 0.36, "learning_rate": 9.54110822164433e-06, "loss": 0.2653, "step": 1154 }, { "epoch": 0.36, "learning_rate": 9.540708141628327e-06, "loss": 0.2887, "step": 1155 }, { "epoch": 0.36, "learning_rate": 9.540308061612324e-06, "loss": 0.4733, "step": 1156 }, { "epoch": 0.36, "learning_rate": 9.53990798159632e-06, "loss": 0.4029, "step": 1157 }, { "epoch": 0.36, "learning_rate": 9.539507901580317e-06, "loss": 0.4267, "step": 1158 }, { "epoch": 0.36, "learning_rate": 9.539107821564314e-06, "loss": 0.3469, "step": 1159 }, { "epoch": 0.36, "learning_rate": 9.53870774154831e-06, "loss": 0.2766, "step": 1160 }, { "epoch": 0.36, "learning_rate": 9.538307661532307e-06, "loss": 0.1703, "step": 1161 }, { "epoch": 0.36, "learning_rate": 9.537907581516304e-06, "loss": 0.4465, "step": 1162 }, { "epoch": 0.36, "learning_rate": 9.5375075015003e-06, "loss": 0.3321, "step": 1163 }, { "epoch": 0.36, "learning_rate": 9.537107421484297e-06, "loss": 0.6842, "step": 1164 }, { "epoch": 0.36, "learning_rate": 9.536707341468294e-06, "loss": 0.2955, "step": 1165 }, { "epoch": 0.36, "learning_rate": 9.53630726145229e-06, "loss": 0.4204, "step": 1166 }, { "epoch": 0.36, "learning_rate": 9.535907181436287e-06, "loss": 0.2922, "step": 1167 }, { "epoch": 0.36, "learning_rate": 9.535507101420286e-06, "loss": 0.2643, "step": 1168 }, { "epoch": 0.36, "learning_rate": 9.535107021404282e-06, "loss": 0.2317, "step": 1169 }, { "epoch": 0.36, "learning_rate": 9.534706941388279e-06, "loss": 0.2758, "step": 1170 }, { "epoch": 0.36, "learning_rate": 9.534306861372276e-06, "loss": 0.2119, "step": 1171 }, { "epoch": 0.36, "learning_rate": 9.533906781356272e-06, "loss": 0.1563, "step": 1172 }, { "epoch": 0.36, "learning_rate": 9.53350670134027e-06, "loss": 0.3464, "step": 1173 }, { "epoch": 0.36, "learning_rate": 9.533106621324266e-06, "loss": 0.2422, "step": 1174 }, { "epoch": 0.36, "learning_rate": 9.532706541308263e-06, "loss": 0.235, "step": 1175 }, { "epoch": 0.37, "learning_rate": 9.53230646129226e-06, "loss": 0.3248, "step": 1176 }, { "epoch": 0.37, "learning_rate": 9.531906381276256e-06, "loss": 0.3686, "step": 1177 }, { "epoch": 0.37, "learning_rate": 9.531506301260253e-06, "loss": 0.3047, "step": 1178 }, { "epoch": 0.37, "learning_rate": 9.53110622124425e-06, "loss": 0.3649, "step": 1179 }, { "epoch": 0.37, "learning_rate": 9.530706141228246e-06, "loss": 0.4476, "step": 1180 }, { "epoch": 0.37, "learning_rate": 9.530306061212243e-06, "loss": 0.3395, "step": 1181 }, { "epoch": 0.37, "learning_rate": 9.529905981196241e-06, "loss": 0.2529, "step": 1182 }, { "epoch": 0.37, "learning_rate": 9.529505901180238e-06, "loss": 0.3006, "step": 1183 }, { "epoch": 0.37, "learning_rate": 9.529105821164234e-06, "loss": 0.1767, "step": 1184 }, { "epoch": 0.37, "learning_rate": 9.528705741148231e-06, "loss": 0.3911, "step": 1185 }, { "epoch": 0.37, "learning_rate": 9.528305661132228e-06, "loss": 0.3144, "step": 1186 }, { "epoch": 0.37, "learning_rate": 9.527905581116223e-06, "loss": 0.2969, "step": 1187 }, { "epoch": 0.37, "learning_rate": 9.527505501100221e-06, "loss": 0.324, "step": 1188 }, { "epoch": 0.37, "learning_rate": 9.527105421084218e-06, "loss": 0.2062, "step": 1189 }, { "epoch": 0.37, "learning_rate": 9.526705341068215e-06, "loss": 0.3515, "step": 1190 }, { "epoch": 0.37, "learning_rate": 9.526305261052211e-06, "loss": 0.5104, "step": 1191 }, { "epoch": 0.37, "learning_rate": 9.525905181036208e-06, "loss": 0.3626, "step": 1192 }, { "epoch": 0.37, "learning_rate": 9.525505101020205e-06, "loss": 0.3732, "step": 1193 }, { "epoch": 0.37, "learning_rate": 9.525105021004201e-06, "loss": 0.2255, "step": 1194 }, { "epoch": 0.37, "learning_rate": 9.524704940988198e-06, "loss": 0.2029, "step": 1195 }, { "epoch": 0.37, "learning_rate": 9.524304860972196e-06, "loss": 0.2813, "step": 1196 }, { "epoch": 0.37, "learning_rate": 9.523904780956193e-06, "loss": 0.2851, "step": 1197 }, { "epoch": 0.37, "learning_rate": 9.52350470094019e-06, "loss": 0.2785, "step": 1198 }, { "epoch": 0.37, "learning_rate": 9.523104620924185e-06, "loss": 0.3082, "step": 1199 }, { "epoch": 0.37, "learning_rate": 9.522704540908181e-06, "loss": 0.317, "step": 1200 }, { "epoch": 0.37, "learning_rate": 9.522304460892178e-06, "loss": 0.3122, "step": 1201 }, { "epoch": 0.37, "learning_rate": 9.521904380876176e-06, "loss": 0.4371, "step": 1202 }, { "epoch": 0.37, "learning_rate": 9.521504300860173e-06, "loss": 0.2604, "step": 1203 }, { "epoch": 0.37, "learning_rate": 9.52110422084417e-06, "loss": 0.4975, "step": 1204 }, { "epoch": 0.37, "learning_rate": 9.520704140828167e-06, "loss": 0.259, "step": 1205 }, { "epoch": 0.37, "learning_rate": 9.520304060812163e-06, "loss": 0.2871, "step": 1206 }, { "epoch": 0.37, "learning_rate": 9.51990398079616e-06, "loss": 0.3652, "step": 1207 }, { "epoch": 0.37, "learning_rate": 9.519503900780157e-06, "loss": 0.353, "step": 1208 }, { "epoch": 0.38, "learning_rate": 9.519103820764153e-06, "loss": 0.2443, "step": 1209 }, { "epoch": 0.38, "learning_rate": 9.518703740748152e-06, "loss": 0.2817, "step": 1210 }, { "epoch": 0.38, "learning_rate": 9.518303660732148e-06, "loss": 0.3088, "step": 1211 }, { "epoch": 0.38, "learning_rate": 9.517903580716143e-06, "loss": 0.4445, "step": 1212 }, { "epoch": 0.38, "learning_rate": 9.51750350070014e-06, "loss": 0.3276, "step": 1213 }, { "epoch": 0.38, "learning_rate": 9.517103420684137e-06, "loss": 0.338, "step": 1214 }, { "epoch": 0.38, "learning_rate": 9.516703340668133e-06, "loss": 0.3492, "step": 1215 }, { "epoch": 0.38, "learning_rate": 9.516303260652132e-06, "loss": 0.3597, "step": 1216 }, { "epoch": 0.38, "learning_rate": 9.515903180636128e-06, "loss": 0.2882, "step": 1217 }, { "epoch": 0.38, "learning_rate": 9.515503100620125e-06, "loss": 0.3467, "step": 1218 }, { "epoch": 0.38, "learning_rate": 9.515103020604122e-06, "loss": 0.2949, "step": 1219 }, { "epoch": 0.38, "learning_rate": 9.514702940588118e-06, "loss": 0.2821, "step": 1220 }, { "epoch": 0.38, "learning_rate": 9.514302860572115e-06, "loss": 0.2696, "step": 1221 }, { "epoch": 0.38, "learning_rate": 9.513902780556112e-06, "loss": 0.2201, "step": 1222 }, { "epoch": 0.38, "learning_rate": 9.513502700540109e-06, "loss": 0.4575, "step": 1223 }, { "epoch": 0.38, "learning_rate": 9.513102620524105e-06, "loss": 0.341, "step": 1224 }, { "epoch": 0.38, "learning_rate": 9.512702540508102e-06, "loss": 0.2408, "step": 1225 }, { "epoch": 0.38, "learning_rate": 9.512302460492099e-06, "loss": 0.2037, "step": 1226 }, { "epoch": 0.38, "learning_rate": 9.511902380476095e-06, "loss": 0.2559, "step": 1227 }, { "epoch": 0.38, "learning_rate": 9.511502300460092e-06, "loss": 0.2509, "step": 1228 }, { "epoch": 0.38, "learning_rate": 9.511102220444089e-06, "loss": 0.4496, "step": 1229 }, { "epoch": 0.38, "learning_rate": 9.510702140428087e-06, "loss": 0.396, "step": 1230 }, { "epoch": 0.38, "learning_rate": 9.510302060412084e-06, "loss": 0.2606, "step": 1231 }, { "epoch": 0.38, "learning_rate": 9.50990198039608e-06, "loss": 0.4067, "step": 1232 }, { "epoch": 0.38, "learning_rate": 9.509501900380077e-06, "loss": 0.2171, "step": 1233 }, { "epoch": 0.38, "learning_rate": 9.509101820364074e-06, "loss": 0.2622, "step": 1234 }, { "epoch": 0.38, "learning_rate": 9.50870174034807e-06, "loss": 0.3751, "step": 1235 }, { "epoch": 0.38, "learning_rate": 9.508301660332067e-06, "loss": 0.426, "step": 1236 }, { "epoch": 0.38, "learning_rate": 9.507901580316064e-06, "loss": 0.3453, "step": 1237 }, { "epoch": 0.38, "learning_rate": 9.50750150030006e-06, "loss": 0.3678, "step": 1238 }, { "epoch": 0.38, "learning_rate": 9.507101420284057e-06, "loss": 0.4165, "step": 1239 }, { "epoch": 0.38, "learning_rate": 9.506701340268054e-06, "loss": 0.2927, "step": 1240 }, { "epoch": 0.39, "learning_rate": 9.50630126025205e-06, "loss": 0.3639, "step": 1241 }, { "epoch": 0.39, "learning_rate": 9.505901180236047e-06, "loss": 0.2385, "step": 1242 }, { "epoch": 0.39, "learning_rate": 9.505501100220044e-06, "loss": 0.3612, "step": 1243 }, { "epoch": 0.39, "learning_rate": 9.505101020204042e-06, "loss": 0.1734, "step": 1244 }, { "epoch": 0.39, "learning_rate": 9.504700940188039e-06, "loss": 0.4239, "step": 1245 }, { "epoch": 0.39, "learning_rate": 9.504300860172036e-06, "loss": 0.3749, "step": 1246 }, { "epoch": 0.39, "learning_rate": 9.503900780156032e-06, "loss": 0.4838, "step": 1247 }, { "epoch": 0.39, "learning_rate": 9.503500700140029e-06, "loss": 0.2575, "step": 1248 }, { "epoch": 0.39, "learning_rate": 9.503100620124026e-06, "loss": 0.1643, "step": 1249 }, { "epoch": 0.39, "learning_rate": 9.502700540108022e-06, "loss": 0.2187, "step": 1250 }, { "epoch": 0.39, "learning_rate": 9.50230046009202e-06, "loss": 0.4334, "step": 1251 }, { "epoch": 0.39, "learning_rate": 9.501900380076016e-06, "loss": 0.3284, "step": 1252 }, { "epoch": 0.39, "learning_rate": 9.501500300060013e-06, "loss": 0.2457, "step": 1253 }, { "epoch": 0.39, "learning_rate": 9.50110022004401e-06, "loss": 0.3465, "step": 1254 }, { "epoch": 0.39, "learning_rate": 9.500700140028006e-06, "loss": 0.3966, "step": 1255 }, { "epoch": 0.39, "learning_rate": 9.500300060012003e-06, "loss": 0.4353, "step": 1256 }, { "epoch": 0.39, "learning_rate": 9.499899979996e-06, "loss": 0.3697, "step": 1257 }, { "epoch": 0.39, "learning_rate": 9.499499899979998e-06, "loss": 0.5286, "step": 1258 }, { "epoch": 0.39, "learning_rate": 9.499099819963994e-06, "loss": 0.236, "step": 1259 }, { "epoch": 0.39, "learning_rate": 9.498699739947991e-06, "loss": 0.3239, "step": 1260 }, { "epoch": 0.39, "learning_rate": 9.498299659931988e-06, "loss": 0.2678, "step": 1261 }, { "epoch": 0.39, "learning_rate": 9.497899579915983e-06, "loss": 0.3999, "step": 1262 }, { "epoch": 0.39, "learning_rate": 9.497499499899981e-06, "loss": 0.3518, "step": 1263 }, { "epoch": 0.39, "learning_rate": 9.497099419883978e-06, "loss": 0.2993, "step": 1264 }, { "epoch": 0.39, "learning_rate": 9.496699339867974e-06, "loss": 0.2129, "step": 1265 }, { "epoch": 0.39, "learning_rate": 9.496299259851971e-06, "loss": 0.267, "step": 1266 }, { "epoch": 0.39, "learning_rate": 9.495899179835968e-06, "loss": 0.3525, "step": 1267 }, { "epoch": 0.39, "learning_rate": 9.495499099819965e-06, "loss": 0.2318, "step": 1268 }, { "epoch": 0.39, "learning_rate": 9.495099019803961e-06, "loss": 0.3086, "step": 1269 }, { "epoch": 0.39, "learning_rate": 9.494698939787958e-06, "loss": 0.2828, "step": 1270 }, { "epoch": 0.39, "learning_rate": 9.494298859771956e-06, "loss": 0.3252, "step": 1271 }, { "epoch": 0.39, "learning_rate": 9.493898779755953e-06, "loss": 0.2957, "step": 1272 }, { "epoch": 0.4, "learning_rate": 9.49349869973995e-06, "loss": 0.3935, "step": 1273 }, { "epoch": 0.4, "learning_rate": 9.493098619723946e-06, "loss": 0.4459, "step": 1274 }, { "epoch": 0.4, "learning_rate": 9.492698539707941e-06, "loss": 0.3218, "step": 1275 }, { "epoch": 0.4, "learning_rate": 9.492298459691938e-06, "loss": 0.2887, "step": 1276 }, { "epoch": 0.4, "learning_rate": 9.491898379675936e-06, "loss": 0.3603, "step": 1277 }, { "epoch": 0.4, "learning_rate": 9.491498299659933e-06, "loss": 0.2276, "step": 1278 }, { "epoch": 0.4, "learning_rate": 9.49109821964393e-06, "loss": 0.3082, "step": 1279 }, { "epoch": 0.4, "learning_rate": 9.490698139627926e-06, "loss": 0.3513, "step": 1280 }, { "epoch": 0.4, "learning_rate": 9.490298059611923e-06, "loss": 0.2673, "step": 1281 }, { "epoch": 0.4, "learning_rate": 9.48989797959592e-06, "loss": 0.3075, "step": 1282 }, { "epoch": 0.4, "learning_rate": 9.489497899579917e-06, "loss": 0.3564, "step": 1283 }, { "epoch": 0.4, "learning_rate": 9.489097819563913e-06, "loss": 0.3187, "step": 1284 }, { "epoch": 0.4, "learning_rate": 9.488697739547912e-06, "loss": 0.3706, "step": 1285 }, { "epoch": 0.4, "learning_rate": 9.488297659531908e-06, "loss": 0.3096, "step": 1286 }, { "epoch": 0.4, "learning_rate": 9.487897579515905e-06, "loss": 0.3799, "step": 1287 }, { "epoch": 0.4, "learning_rate": 9.4874974994999e-06, "loss": 0.3075, "step": 1288 }, { "epoch": 0.4, "learning_rate": 9.487097419483897e-06, "loss": 0.2822, "step": 1289 }, { "epoch": 0.4, "learning_rate": 9.486697339467893e-06, "loss": 0.4497, "step": 1290 }, { "epoch": 0.4, "learning_rate": 9.486297259451892e-06, "loss": 0.4205, "step": 1291 }, { "epoch": 0.4, "learning_rate": 9.485897179435888e-06, "loss": 0.3129, "step": 1292 }, { "epoch": 0.4, "learning_rate": 9.485497099419885e-06, "loss": 0.2616, "step": 1293 }, { "epoch": 0.4, "learning_rate": 9.485097019403882e-06, "loss": 0.3882, "step": 1294 }, { "epoch": 0.4, "learning_rate": 9.484696939387878e-06, "loss": 0.361, "step": 1295 }, { "epoch": 0.4, "learning_rate": 9.484296859371875e-06, "loss": 0.4587, "step": 1296 }, { "epoch": 0.4, "learning_rate": 9.483896779355872e-06, "loss": 0.4336, "step": 1297 }, { "epoch": 0.4, "learning_rate": 9.483496699339868e-06, "loss": 0.3797, "step": 1298 }, { "epoch": 0.4, "learning_rate": 9.483096619323867e-06, "loss": 0.4011, "step": 1299 }, { "epoch": 0.4, "learning_rate": 9.482696539307862e-06, "loss": 0.354, "step": 1300 }, { "epoch": 0.4, "learning_rate": 9.482296459291859e-06, "loss": 0.3429, "step": 1301 }, { "epoch": 0.4, "learning_rate": 9.481896379275855e-06, "loss": 0.2402, "step": 1302 }, { "epoch": 0.4, "learning_rate": 9.481496299259852e-06, "loss": 0.5506, "step": 1303 }, { "epoch": 0.4, "learning_rate": 9.481096219243849e-06, "loss": 0.414, "step": 1304 }, { "epoch": 0.41, "learning_rate": 9.480696139227847e-06, "loss": 0.1906, "step": 1305 }, { "epoch": 0.41, "learning_rate": 9.480296059211844e-06, "loss": 0.3148, "step": 1306 }, { "epoch": 0.41, "learning_rate": 9.47989597919584e-06, "loss": 0.2886, "step": 1307 }, { "epoch": 0.41, "learning_rate": 9.479495899179837e-06, "loss": 0.271, "step": 1308 }, { "epoch": 0.41, "learning_rate": 9.479095819163834e-06, "loss": 0.403, "step": 1309 }, { "epoch": 0.41, "learning_rate": 9.47869573914783e-06, "loss": 0.2368, "step": 1310 }, { "epoch": 0.41, "learning_rate": 9.478295659131827e-06, "loss": 0.486, "step": 1311 }, { "epoch": 0.41, "learning_rate": 9.477895579115824e-06, "loss": 0.337, "step": 1312 }, { "epoch": 0.41, "learning_rate": 9.47749549909982e-06, "loss": 0.3997, "step": 1313 }, { "epoch": 0.41, "learning_rate": 9.477095419083817e-06, "loss": 0.2758, "step": 1314 }, { "epoch": 0.41, "learning_rate": 9.476695339067814e-06, "loss": 0.3146, "step": 1315 }, { "epoch": 0.41, "learning_rate": 9.47629525905181e-06, "loss": 0.4778, "step": 1316 }, { "epoch": 0.41, "learning_rate": 9.475895179035807e-06, "loss": 0.2324, "step": 1317 }, { "epoch": 0.41, "learning_rate": 9.475495099019804e-06, "loss": 0.3506, "step": 1318 }, { "epoch": 0.41, "learning_rate": 9.475095019003802e-06, "loss": 0.2331, "step": 1319 }, { "epoch": 0.41, "learning_rate": 9.474694938987799e-06, "loss": 0.2975, "step": 1320 }, { "epoch": 0.41, "learning_rate": 9.474294858971796e-06, "loss": 0.1428, "step": 1321 }, { "epoch": 0.41, "learning_rate": 9.473894778955792e-06, "loss": 0.2796, "step": 1322 }, { "epoch": 0.41, "learning_rate": 9.473494698939789e-06, "loss": 0.2444, "step": 1323 }, { "epoch": 0.41, "learning_rate": 9.473094618923786e-06, "loss": 0.4216, "step": 1324 }, { "epoch": 0.41, "learning_rate": 9.472694538907782e-06, "loss": 0.3413, "step": 1325 }, { "epoch": 0.41, "learning_rate": 9.472294458891779e-06, "loss": 0.2288, "step": 1326 }, { "epoch": 0.41, "learning_rate": 9.471894378875776e-06, "loss": 0.4084, "step": 1327 }, { "epoch": 0.41, "learning_rate": 9.471494298859772e-06, "loss": 0.3647, "step": 1328 }, { "epoch": 0.41, "learning_rate": 9.47109421884377e-06, "loss": 0.3022, "step": 1329 }, { "epoch": 0.41, "learning_rate": 9.470694138827766e-06, "loss": 0.2764, "step": 1330 }, { "epoch": 0.41, "learning_rate": 9.470294058811763e-06, "loss": 0.4262, "step": 1331 }, { "epoch": 0.41, "learning_rate": 9.46989397879576e-06, "loss": 0.4444, "step": 1332 }, { "epoch": 0.41, "learning_rate": 9.469493898779758e-06, "loss": 0.2194, "step": 1333 }, { "epoch": 0.41, "learning_rate": 9.469093818763754e-06, "loss": 0.2759, "step": 1334 }, { "epoch": 0.41, "learning_rate": 9.468693738747751e-06, "loss": 0.3288, "step": 1335 }, { "epoch": 0.41, "learning_rate": 9.468293658731748e-06, "loss": 0.2736, "step": 1336 }, { "epoch": 0.41, "learning_rate": 9.467893578715744e-06, "loss": 0.3062, "step": 1337 }, { "epoch": 0.42, "learning_rate": 9.467493498699741e-06, "loss": 0.2543, "step": 1338 }, { "epoch": 0.42, "learning_rate": 9.467093418683738e-06, "loss": 0.3645, "step": 1339 }, { "epoch": 0.42, "learning_rate": 9.466693338667734e-06, "loss": 0.461, "step": 1340 }, { "epoch": 0.42, "learning_rate": 9.466293258651731e-06, "loss": 0.231, "step": 1341 }, { "epoch": 0.42, "learning_rate": 9.465893178635728e-06, "loss": 0.2851, "step": 1342 }, { "epoch": 0.42, "learning_rate": 9.465493098619724e-06, "loss": 0.4469, "step": 1343 }, { "epoch": 0.42, "learning_rate": 9.465093018603721e-06, "loss": 0.1897, "step": 1344 }, { "epoch": 0.42, "learning_rate": 9.464692938587718e-06, "loss": 0.4121, "step": 1345 }, { "epoch": 0.42, "learning_rate": 9.464292858571715e-06, "loss": 0.3273, "step": 1346 }, { "epoch": 0.42, "learning_rate": 9.463892778555713e-06, "loss": 0.3019, "step": 1347 }, { "epoch": 0.42, "learning_rate": 9.46349269853971e-06, "loss": 0.3132, "step": 1348 }, { "epoch": 0.42, "learning_rate": 9.463092618523706e-06, "loss": 0.2928, "step": 1349 }, { "epoch": 0.42, "learning_rate": 9.462692538507703e-06, "loss": 0.2723, "step": 1350 }, { "epoch": 0.42, "learning_rate": 9.462292458491698e-06, "loss": 0.4156, "step": 1351 }, { "epoch": 0.42, "learning_rate": 9.461892378475696e-06, "loss": 0.2091, "step": 1352 }, { "epoch": 0.42, "learning_rate": 9.461492298459693e-06, "loss": 0.2615, "step": 1353 }, { "epoch": 0.42, "learning_rate": 9.46109221844369e-06, "loss": 0.289, "step": 1354 }, { "epoch": 0.42, "learning_rate": 9.460692138427686e-06, "loss": 0.3844, "step": 1355 }, { "epoch": 0.42, "learning_rate": 9.460292058411683e-06, "loss": 0.3217, "step": 1356 }, { "epoch": 0.42, "learning_rate": 9.45989197839568e-06, "loss": 0.294, "step": 1357 }, { "epoch": 0.42, "learning_rate": 9.459491898379676e-06, "loss": 0.3141, "step": 1358 }, { "epoch": 0.42, "learning_rate": 9.459091818363673e-06, "loss": 0.3605, "step": 1359 }, { "epoch": 0.42, "learning_rate": 9.45869173834767e-06, "loss": 0.2562, "step": 1360 }, { "epoch": 0.42, "learning_rate": 9.458291658331668e-06, "loss": 0.3461, "step": 1361 }, { "epoch": 0.42, "learning_rate": 9.457891578315665e-06, "loss": 0.2419, "step": 1362 }, { "epoch": 0.42, "learning_rate": 9.45749149829966e-06, "loss": 0.2239, "step": 1363 }, { "epoch": 0.42, "learning_rate": 9.457091418283657e-06, "loss": 0.4246, "step": 1364 }, { "epoch": 0.42, "learning_rate": 9.456691338267653e-06, "loss": 0.3855, "step": 1365 }, { "epoch": 0.42, "learning_rate": 9.456291258251652e-06, "loss": 0.3162, "step": 1366 }, { "epoch": 0.42, "learning_rate": 9.455891178235648e-06, "loss": 0.1684, "step": 1367 }, { "epoch": 0.42, "learning_rate": 9.455491098219645e-06, "loss": 0.423, "step": 1368 }, { "epoch": 0.42, "learning_rate": 9.455091018203642e-06, "loss": 0.1562, "step": 1369 }, { "epoch": 0.43, "learning_rate": 9.454690938187638e-06, "loss": 0.5017, "step": 1370 }, { "epoch": 0.43, "learning_rate": 9.454290858171635e-06, "loss": 0.186, "step": 1371 }, { "epoch": 0.43, "learning_rate": 9.453890778155632e-06, "loss": 0.293, "step": 1372 }, { "epoch": 0.43, "learning_rate": 9.453490698139628e-06, "loss": 0.4248, "step": 1373 }, { "epoch": 0.43, "learning_rate": 9.453090618123625e-06, "loss": 0.2581, "step": 1374 }, { "epoch": 0.43, "learning_rate": 9.452690538107623e-06, "loss": 0.3134, "step": 1375 }, { "epoch": 0.43, "learning_rate": 9.452290458091618e-06, "loss": 0.4012, "step": 1376 }, { "epoch": 0.43, "learning_rate": 9.451890378075615e-06, "loss": 0.2602, "step": 1377 }, { "epoch": 0.43, "learning_rate": 9.451490298059612e-06, "loss": 0.3033, "step": 1378 }, { "epoch": 0.43, "learning_rate": 9.451090218043609e-06, "loss": 0.326, "step": 1379 }, { "epoch": 0.43, "learning_rate": 9.450690138027607e-06, "loss": 0.3096, "step": 1380 }, { "epoch": 0.43, "learning_rate": 9.450290058011604e-06, "loss": 0.2581, "step": 1381 }, { "epoch": 0.43, "learning_rate": 9.4498899779956e-06, "loss": 0.2122, "step": 1382 }, { "epoch": 0.43, "learning_rate": 9.449489897979597e-06, "loss": 0.3236, "step": 1383 }, { "epoch": 0.43, "learning_rate": 9.449089817963594e-06, "loss": 0.3525, "step": 1384 }, { "epoch": 0.43, "learning_rate": 9.44868973794759e-06, "loss": 0.335, "step": 1385 }, { "epoch": 0.43, "learning_rate": 9.448289657931587e-06, "loss": 0.3375, "step": 1386 }, { "epoch": 0.43, "learning_rate": 9.447889577915584e-06, "loss": 0.2555, "step": 1387 }, { "epoch": 0.43, "learning_rate": 9.44748949789958e-06, "loss": 0.3404, "step": 1388 }, { "epoch": 0.43, "learning_rate": 9.447089417883577e-06, "loss": 0.3915, "step": 1389 }, { "epoch": 0.43, "learning_rate": 9.446689337867574e-06, "loss": 0.3, "step": 1390 }, { "epoch": 0.43, "learning_rate": 9.44628925785157e-06, "loss": 0.3552, "step": 1391 }, { "epoch": 0.43, "learning_rate": 9.445889177835567e-06, "loss": 0.4283, "step": 1392 }, { "epoch": 0.43, "learning_rate": 9.445489097819564e-06, "loss": 0.3311, "step": 1393 }, { "epoch": 0.43, "learning_rate": 9.445089017803562e-06, "loss": 0.348, "step": 1394 }, { "epoch": 0.43, "learning_rate": 9.444688937787559e-06, "loss": 0.3527, "step": 1395 }, { "epoch": 0.43, "learning_rate": 9.444288857771556e-06, "loss": 0.3652, "step": 1396 }, { "epoch": 0.43, "learning_rate": 9.443888777755552e-06, "loss": 0.3347, "step": 1397 }, { "epoch": 0.43, "learning_rate": 9.443488697739549e-06, "loss": 0.2987, "step": 1398 }, { "epoch": 0.43, "learning_rate": 9.443088617723546e-06, "loss": 0.1279, "step": 1399 }, { "epoch": 0.43, "learning_rate": 9.442688537707542e-06, "loss": 0.4188, "step": 1400 }, { "epoch": 0.43, "learning_rate": 9.442288457691539e-06, "loss": 0.2865, "step": 1401 }, { "epoch": 0.44, "learning_rate": 9.441888377675536e-06, "loss": 0.2455, "step": 1402 }, { "epoch": 0.44, "learning_rate": 9.441488297659532e-06, "loss": 0.2543, "step": 1403 }, { "epoch": 0.44, "learning_rate": 9.441088217643529e-06, "loss": 0.4182, "step": 1404 }, { "epoch": 0.44, "learning_rate": 9.440688137627526e-06, "loss": 0.348, "step": 1405 }, { "epoch": 0.44, "learning_rate": 9.440288057611522e-06, "loss": 0.2494, "step": 1406 }, { "epoch": 0.44, "learning_rate": 9.43988797759552e-06, "loss": 0.5064, "step": 1407 }, { "epoch": 0.44, "learning_rate": 9.439487897579518e-06, "loss": 0.2429, "step": 1408 }, { "epoch": 0.44, "learning_rate": 9.439087817563514e-06, "loss": 0.1936, "step": 1409 }, { "epoch": 0.44, "learning_rate": 9.438687737547511e-06, "loss": 0.5248, "step": 1410 }, { "epoch": 0.44, "learning_rate": 9.438287657531508e-06, "loss": 0.2448, "step": 1411 }, { "epoch": 0.44, "learning_rate": 9.437887577515504e-06, "loss": 0.3051, "step": 1412 }, { "epoch": 0.44, "learning_rate": 9.437487497499501e-06, "loss": 0.5711, "step": 1413 }, { "epoch": 0.44, "learning_rate": 9.437087417483498e-06, "loss": 0.2194, "step": 1414 }, { "epoch": 0.44, "learning_rate": 9.436687337467494e-06, "loss": 0.3579, "step": 1415 }, { "epoch": 0.44, "learning_rate": 9.436287257451491e-06, "loss": 0.5073, "step": 1416 }, { "epoch": 0.44, "learning_rate": 9.435887177435488e-06, "loss": 0.2535, "step": 1417 }, { "epoch": 0.44, "learning_rate": 9.435487097419484e-06, "loss": 0.4457, "step": 1418 }, { "epoch": 0.44, "learning_rate": 9.435087017403481e-06, "loss": 0.3004, "step": 1419 }, { "epoch": 0.44, "learning_rate": 9.434686937387478e-06, "loss": 0.2345, "step": 1420 }, { "epoch": 0.44, "learning_rate": 9.434286857371474e-06, "loss": 0.2967, "step": 1421 }, { "epoch": 0.44, "learning_rate": 9.433886777355473e-06, "loss": 0.3683, "step": 1422 }, { "epoch": 0.44, "learning_rate": 9.43348669733947e-06, "loss": 0.1915, "step": 1423 }, { "epoch": 0.44, "learning_rate": 9.433086617323466e-06, "loss": 0.3945, "step": 1424 }, { "epoch": 0.44, "learning_rate": 9.432686537307463e-06, "loss": 0.5063, "step": 1425 }, { "epoch": 0.44, "learning_rate": 9.432286457291458e-06, "loss": 0.2297, "step": 1426 }, { "epoch": 0.44, "learning_rate": 9.431886377275455e-06, "loss": 0.2759, "step": 1427 }, { "epoch": 0.44, "learning_rate": 9.431486297259453e-06, "loss": 0.3737, "step": 1428 }, { "epoch": 0.44, "learning_rate": 9.43108621724345e-06, "loss": 0.1634, "step": 1429 }, { "epoch": 0.44, "learning_rate": 9.430686137227446e-06, "loss": 0.277, "step": 1430 }, { "epoch": 0.44, "learning_rate": 9.430286057211443e-06, "loss": 0.1775, "step": 1431 }, { "epoch": 0.44, "learning_rate": 9.42988597719544e-06, "loss": 0.2234, "step": 1432 }, { "epoch": 0.44, "learning_rate": 9.429485897179436e-06, "loss": 0.3704, "step": 1433 }, { "epoch": 0.45, "learning_rate": 9.429085817163433e-06, "loss": 0.5401, "step": 1434 }, { "epoch": 0.45, "learning_rate": 9.42868573714743e-06, "loss": 0.2932, "step": 1435 }, { "epoch": 0.45, "learning_rate": 9.428285657131428e-06, "loss": 0.5358, "step": 1436 }, { "epoch": 0.45, "learning_rate": 9.427885577115425e-06, "loss": 0.3148, "step": 1437 }, { "epoch": 0.45, "learning_rate": 9.427485497099422e-06, "loss": 0.2369, "step": 1438 }, { "epoch": 0.45, "learning_rate": 9.427085417083417e-06, "loss": 0.3434, "step": 1439 }, { "epoch": 0.45, "learning_rate": 9.426685337067413e-06, "loss": 0.3046, "step": 1440 }, { "epoch": 0.45, "learning_rate": 9.426285257051412e-06, "loss": 0.337, "step": 1441 }, { "epoch": 0.45, "learning_rate": 9.425885177035408e-06, "loss": 0.3454, "step": 1442 }, { "epoch": 0.45, "learning_rate": 9.425485097019405e-06, "loss": 0.3564, "step": 1443 }, { "epoch": 0.45, "learning_rate": 9.425085017003402e-06, "loss": 0.3259, "step": 1444 }, { "epoch": 0.45, "learning_rate": 9.424684936987398e-06, "loss": 0.2233, "step": 1445 }, { "epoch": 0.45, "learning_rate": 9.424284856971395e-06, "loss": 0.2402, "step": 1446 }, { "epoch": 0.45, "learning_rate": 9.423884776955392e-06, "loss": 0.2476, "step": 1447 }, { "epoch": 0.45, "learning_rate": 9.423484696939388e-06, "loss": 0.3784, "step": 1448 }, { "epoch": 0.45, "learning_rate": 9.423084616923385e-06, "loss": 0.1844, "step": 1449 }, { "epoch": 0.45, "learning_rate": 9.422684536907383e-06, "loss": 0.3763, "step": 1450 }, { "epoch": 0.45, "learning_rate": 9.422284456891378e-06, "loss": 0.425, "step": 1451 }, { "epoch": 0.45, "learning_rate": 9.421884376875375e-06, "loss": 0.1952, "step": 1452 }, { "epoch": 0.45, "learning_rate": 9.421484296859372e-06, "loss": 0.3189, "step": 1453 }, { "epoch": 0.45, "learning_rate": 9.421084216843368e-06, "loss": 0.3691, "step": 1454 }, { "epoch": 0.45, "learning_rate": 9.420684136827367e-06, "loss": 0.3288, "step": 1455 }, { "epoch": 0.45, "learning_rate": 9.420284056811364e-06, "loss": 0.2324, "step": 1456 }, { "epoch": 0.45, "learning_rate": 9.41988397679536e-06, "loss": 0.4346, "step": 1457 }, { "epoch": 0.45, "learning_rate": 9.419483896779357e-06, "loss": 0.2671, "step": 1458 }, { "epoch": 0.45, "learning_rate": 9.419083816763354e-06, "loss": 0.3135, "step": 1459 }, { "epoch": 0.45, "learning_rate": 9.41868373674735e-06, "loss": 0.3095, "step": 1460 }, { "epoch": 0.45, "learning_rate": 9.418283656731347e-06, "loss": 0.2096, "step": 1461 }, { "epoch": 0.45, "learning_rate": 9.417883576715344e-06, "loss": 0.3272, "step": 1462 }, { "epoch": 0.45, "learning_rate": 9.41748349669934e-06, "loss": 0.1593, "step": 1463 }, { "epoch": 0.45, "learning_rate": 9.417083416683337e-06, "loss": 0.2981, "step": 1464 }, { "epoch": 0.45, "learning_rate": 9.416683336667334e-06, "loss": 0.419, "step": 1465 }, { "epoch": 0.46, "learning_rate": 9.41628325665133e-06, "loss": 0.2573, "step": 1466 }, { "epoch": 0.46, "learning_rate": 9.415883176635327e-06, "loss": 0.337, "step": 1467 }, { "epoch": 0.46, "learning_rate": 9.415483096619324e-06, "loss": 0.2866, "step": 1468 }, { "epoch": 0.46, "learning_rate": 9.415083016603322e-06, "loss": 0.2435, "step": 1469 }, { "epoch": 0.46, "learning_rate": 9.414682936587319e-06, "loss": 0.2768, "step": 1470 }, { "epoch": 0.46, "learning_rate": 9.414282856571316e-06, "loss": 0.1954, "step": 1471 }, { "epoch": 0.46, "learning_rate": 9.413882776555312e-06, "loss": 0.2233, "step": 1472 }, { "epoch": 0.46, "learning_rate": 9.413482696539309e-06, "loss": 0.3783, "step": 1473 }, { "epoch": 0.46, "learning_rate": 9.413082616523306e-06, "loss": 0.3835, "step": 1474 }, { "epoch": 0.46, "learning_rate": 9.412682536507302e-06, "loss": 0.1517, "step": 1475 }, { "epoch": 0.46, "learning_rate": 9.412282456491299e-06, "loss": 0.2435, "step": 1476 }, { "epoch": 0.46, "learning_rate": 9.411882376475296e-06, "loss": 0.2651, "step": 1477 }, { "epoch": 0.46, "learning_rate": 9.411482296459292e-06, "loss": 0.2501, "step": 1478 }, { "epoch": 0.46, "learning_rate": 9.411082216443289e-06, "loss": 0.1866, "step": 1479 }, { "epoch": 0.46, "learning_rate": 9.410682136427286e-06, "loss": 0.3344, "step": 1480 }, { "epoch": 0.46, "learning_rate": 9.410282056411282e-06, "loss": 0.2184, "step": 1481 }, { "epoch": 0.46, "learning_rate": 9.409881976395279e-06, "loss": 0.3647, "step": 1482 }, { "epoch": 0.46, "learning_rate": 9.409481896379277e-06, "loss": 0.3035, "step": 1483 }, { "epoch": 0.46, "learning_rate": 9.409081816363274e-06, "loss": 0.4354, "step": 1484 }, { "epoch": 0.46, "learning_rate": 9.408681736347271e-06, "loss": 0.281, "step": 1485 }, { "epoch": 0.46, "learning_rate": 9.408281656331268e-06, "loss": 0.2374, "step": 1486 }, { "epoch": 0.46, "learning_rate": 9.407881576315264e-06, "loss": 0.233, "step": 1487 }, { "epoch": 0.46, "learning_rate": 9.407481496299261e-06, "loss": 0.458, "step": 1488 }, { "epoch": 0.46, "learning_rate": 9.407081416283258e-06, "loss": 0.3064, "step": 1489 }, { "epoch": 0.46, "learning_rate": 9.406681336267254e-06, "loss": 0.3494, "step": 1490 }, { "epoch": 0.46, "learning_rate": 9.406281256251251e-06, "loss": 0.6034, "step": 1491 }, { "epoch": 0.46, "learning_rate": 9.405881176235248e-06, "loss": 0.2319, "step": 1492 }, { "epoch": 0.46, "learning_rate": 9.405481096219244e-06, "loss": 0.2769, "step": 1493 }, { "epoch": 0.46, "learning_rate": 9.405081016203241e-06, "loss": 0.3419, "step": 1494 }, { "epoch": 0.46, "learning_rate": 9.404680936187238e-06, "loss": 0.3558, "step": 1495 }, { "epoch": 0.46, "learning_rate": 9.404280856171234e-06, "loss": 0.3629, "step": 1496 }, { "epoch": 0.46, "learning_rate": 9.403880776155233e-06, "loss": 0.3378, "step": 1497 }, { "epoch": 0.46, "learning_rate": 9.40348069613923e-06, "loss": 0.307, "step": 1498 }, { "epoch": 0.47, "learning_rate": 9.403080616123226e-06, "loss": 0.2099, "step": 1499 }, { "epoch": 0.47, "learning_rate": 9.402680536107223e-06, "loss": 0.1811, "step": 1500 }, { "epoch": 0.47, "learning_rate": 9.40228045609122e-06, "loss": 0.2334, "step": 1501 }, { "epoch": 0.47, "learning_rate": 9.401880376075215e-06, "loss": 0.2486, "step": 1502 }, { "epoch": 0.47, "learning_rate": 9.401480296059213e-06, "loss": 0.3624, "step": 1503 }, { "epoch": 0.47, "learning_rate": 9.40108021604321e-06, "loss": 0.2052, "step": 1504 }, { "epoch": 0.47, "learning_rate": 9.400680136027206e-06, "loss": 0.2343, "step": 1505 }, { "epoch": 0.47, "learning_rate": 9.400280056011203e-06, "loss": 0.321, "step": 1506 }, { "epoch": 0.47, "learning_rate": 9.3998799759952e-06, "loss": 0.3082, "step": 1507 }, { "epoch": 0.47, "learning_rate": 9.399479895979196e-06, "loss": 0.3911, "step": 1508 }, { "epoch": 0.47, "learning_rate": 9.399079815963193e-06, "loss": 0.2312, "step": 1509 }, { "epoch": 0.47, "learning_rate": 9.39867973594719e-06, "loss": 0.3092, "step": 1510 }, { "epoch": 0.47, "learning_rate": 9.398279655931188e-06, "loss": 0.2893, "step": 1511 }, { "epoch": 0.47, "learning_rate": 9.397879575915185e-06, "loss": 0.2937, "step": 1512 }, { "epoch": 0.47, "learning_rate": 9.397479495899181e-06, "loss": 0.5141, "step": 1513 }, { "epoch": 0.47, "learning_rate": 9.397079415883178e-06, "loss": 0.287, "step": 1514 }, { "epoch": 0.47, "learning_rate": 9.396679335867173e-06, "loss": 0.3334, "step": 1515 }, { "epoch": 0.47, "learning_rate": 9.39627925585117e-06, "loss": 0.2754, "step": 1516 }, { "epoch": 0.47, "learning_rate": 9.395879175835168e-06, "loss": 0.2938, "step": 1517 }, { "epoch": 0.47, "learning_rate": 9.395479095819165e-06, "loss": 0.3013, "step": 1518 }, { "epoch": 0.47, "learning_rate": 9.395079015803162e-06, "loss": 0.2013, "step": 1519 }, { "epoch": 0.47, "learning_rate": 9.394678935787158e-06, "loss": 0.2935, "step": 1520 }, { "epoch": 0.47, "learning_rate": 9.394278855771155e-06, "loss": 0.3615, "step": 1521 }, { "epoch": 0.47, "learning_rate": 9.393878775755152e-06, "loss": 0.2923, "step": 1522 }, { "epoch": 0.47, "learning_rate": 9.393478695739148e-06, "loss": 0.3945, "step": 1523 }, { "epoch": 0.47, "learning_rate": 9.393078615723145e-06, "loss": 0.3253, "step": 1524 }, { "epoch": 0.47, "learning_rate": 9.392678535707143e-06, "loss": 0.2023, "step": 1525 }, { "epoch": 0.47, "learning_rate": 9.39227845569114e-06, "loss": 0.1603, "step": 1526 }, { "epoch": 0.47, "learning_rate": 9.391878375675135e-06, "loss": 0.3414, "step": 1527 }, { "epoch": 0.47, "learning_rate": 9.391478295659132e-06, "loss": 0.2871, "step": 1528 }, { "epoch": 0.47, "learning_rate": 9.391078215643128e-06, "loss": 0.3023, "step": 1529 }, { "epoch": 0.47, "learning_rate": 9.390678135627125e-06, "loss": 0.3765, "step": 1530 }, { "epoch": 0.48, "learning_rate": 9.390278055611123e-06, "loss": 0.2512, "step": 1531 }, { "epoch": 0.48, "learning_rate": 9.38987797559512e-06, "loss": 0.3505, "step": 1532 }, { "epoch": 0.48, "learning_rate": 9.389477895579117e-06, "loss": 0.2673, "step": 1533 }, { "epoch": 0.48, "learning_rate": 9.389077815563114e-06, "loss": 0.2807, "step": 1534 }, { "epoch": 0.48, "learning_rate": 9.38867773554711e-06, "loss": 0.3332, "step": 1535 }, { "epoch": 0.48, "learning_rate": 9.388277655531107e-06, "loss": 0.308, "step": 1536 }, { "epoch": 0.48, "learning_rate": 9.387877575515104e-06, "loss": 0.2085, "step": 1537 }, { "epoch": 0.48, "learning_rate": 9.3874774954991e-06, "loss": 0.174, "step": 1538 }, { "epoch": 0.48, "learning_rate": 9.387077415483099e-06, "loss": 0.3531, "step": 1539 }, { "epoch": 0.48, "learning_rate": 9.386677335467094e-06, "loss": 0.3264, "step": 1540 }, { "epoch": 0.48, "learning_rate": 9.38627725545109e-06, "loss": 0.2299, "step": 1541 }, { "epoch": 0.48, "learning_rate": 9.385877175435087e-06, "loss": 0.3616, "step": 1542 }, { "epoch": 0.48, "learning_rate": 9.385477095419084e-06, "loss": 0.1672, "step": 1543 }, { "epoch": 0.48, "learning_rate": 9.38507701540308e-06, "loss": 0.2283, "step": 1544 }, { "epoch": 0.48, "learning_rate": 9.384676935387079e-06, "loss": 0.2454, "step": 1545 }, { "epoch": 0.48, "learning_rate": 9.384276855371075e-06, "loss": 0.3819, "step": 1546 }, { "epoch": 0.48, "learning_rate": 9.383876775355072e-06, "loss": 0.2002, "step": 1547 }, { "epoch": 0.48, "learning_rate": 9.383476695339069e-06, "loss": 0.3059, "step": 1548 }, { "epoch": 0.48, "learning_rate": 9.383076615323066e-06, "loss": 0.3174, "step": 1549 }, { "epoch": 0.48, "learning_rate": 9.382676535307062e-06, "loss": 0.3218, "step": 1550 }, { "epoch": 0.48, "learning_rate": 9.382276455291059e-06, "loss": 0.3653, "step": 1551 }, { "epoch": 0.48, "learning_rate": 9.381876375275056e-06, "loss": 0.3343, "step": 1552 }, { "epoch": 0.48, "learning_rate": 9.381476295259052e-06, "loss": 0.2054, "step": 1553 }, { "epoch": 0.48, "learning_rate": 9.381076215243049e-06, "loss": 0.2954, "step": 1554 }, { "epoch": 0.48, "learning_rate": 9.380676135227046e-06, "loss": 0.4292, "step": 1555 }, { "epoch": 0.48, "learning_rate": 9.380276055211042e-06, "loss": 0.2548, "step": 1556 }, { "epoch": 0.48, "learning_rate": 9.379875975195039e-06, "loss": 0.2307, "step": 1557 }, { "epoch": 0.48, "learning_rate": 9.379475895179036e-06, "loss": 0.2846, "step": 1558 }, { "epoch": 0.48, "learning_rate": 9.379075815163034e-06, "loss": 0.2725, "step": 1559 }, { "epoch": 0.48, "learning_rate": 9.37867573514703e-06, "loss": 0.3035, "step": 1560 }, { "epoch": 0.48, "learning_rate": 9.378275655131027e-06, "loss": 0.3665, "step": 1561 }, { "epoch": 0.48, "learning_rate": 9.377875575115024e-06, "loss": 0.2334, "step": 1562 }, { "epoch": 0.49, "learning_rate": 9.377475495099021e-06, "loss": 0.2663, "step": 1563 }, { "epoch": 0.49, "learning_rate": 9.377075415083018e-06, "loss": 0.3689, "step": 1564 }, { "epoch": 0.49, "learning_rate": 9.376675335067014e-06, "loss": 0.3757, "step": 1565 }, { "epoch": 0.49, "learning_rate": 9.376275255051011e-06, "loss": 0.1945, "step": 1566 }, { "epoch": 0.49, "learning_rate": 9.375875175035008e-06, "loss": 0.1933, "step": 1567 }, { "epoch": 0.49, "learning_rate": 9.375475095019004e-06, "loss": 0.2611, "step": 1568 }, { "epoch": 0.49, "learning_rate": 9.375075015003001e-06, "loss": 0.1696, "step": 1569 }, { "epoch": 0.49, "learning_rate": 9.374674934986998e-06, "loss": 0.2558, "step": 1570 }, { "epoch": 0.49, "learning_rate": 9.374274854970994e-06, "loss": 0.3061, "step": 1571 }, { "epoch": 0.49, "learning_rate": 9.373874774954993e-06, "loss": 0.3159, "step": 1572 }, { "epoch": 0.49, "learning_rate": 9.37347469493899e-06, "loss": 0.2561, "step": 1573 }, { "epoch": 0.49, "learning_rate": 9.373074614922986e-06, "loss": 0.2541, "step": 1574 }, { "epoch": 0.49, "learning_rate": 9.372674534906983e-06, "loss": 0.3345, "step": 1575 }, { "epoch": 0.49, "learning_rate": 9.37227445489098e-06, "loss": 0.2735, "step": 1576 }, { "epoch": 0.49, "learning_rate": 9.371874374874976e-06, "loss": 0.3253, "step": 1577 }, { "epoch": 0.49, "learning_rate": 9.371474294858973e-06, "loss": 0.4633, "step": 1578 }, { "epoch": 0.49, "learning_rate": 9.37107421484297e-06, "loss": 0.3432, "step": 1579 }, { "epoch": 0.49, "learning_rate": 9.370674134826966e-06, "loss": 0.2525, "step": 1580 }, { "epoch": 0.49, "learning_rate": 9.370274054810963e-06, "loss": 0.1503, "step": 1581 }, { "epoch": 0.49, "learning_rate": 9.36987397479496e-06, "loss": 0.1807, "step": 1582 }, { "epoch": 0.49, "learning_rate": 9.369473894778956e-06, "loss": 0.3151, "step": 1583 }, { "epoch": 0.49, "learning_rate": 9.369073814762953e-06, "loss": 0.4083, "step": 1584 }, { "epoch": 0.49, "learning_rate": 9.36867373474695e-06, "loss": 0.2268, "step": 1585 }, { "epoch": 0.49, "learning_rate": 9.368273654730948e-06, "loss": 0.2303, "step": 1586 }, { "epoch": 0.49, "learning_rate": 9.367873574714945e-06, "loss": 0.1413, "step": 1587 }, { "epoch": 0.49, "learning_rate": 9.367473494698941e-06, "loss": 0.3891, "step": 1588 }, { "epoch": 0.49, "learning_rate": 9.367073414682938e-06, "loss": 0.2606, "step": 1589 }, { "epoch": 0.49, "learning_rate": 9.366673334666933e-06, "loss": 0.2502, "step": 1590 }, { "epoch": 0.49, "learning_rate": 9.36627325465093e-06, "loss": 0.3243, "step": 1591 }, { "epoch": 0.49, "learning_rate": 9.365873174634928e-06, "loss": 0.4502, "step": 1592 }, { "epoch": 0.49, "learning_rate": 9.365473094618925e-06, "loss": 0.2354, "step": 1593 }, { "epoch": 0.49, "learning_rate": 9.365073014602922e-06, "loss": 0.3903, "step": 1594 }, { "epoch": 0.5, "learning_rate": 9.364672934586918e-06, "loss": 0.2727, "step": 1595 }, { "epoch": 0.5, "learning_rate": 9.364272854570915e-06, "loss": 0.2421, "step": 1596 }, { "epoch": 0.5, "learning_rate": 9.363872774554912e-06, "loss": 0.2151, "step": 1597 }, { "epoch": 0.5, "learning_rate": 9.363472694538908e-06, "loss": 0.2317, "step": 1598 }, { "epoch": 0.5, "learning_rate": 9.363072614522905e-06, "loss": 0.3364, "step": 1599 }, { "epoch": 0.5, "learning_rate": 9.362672534506903e-06, "loss": 0.4368, "step": 1600 }, { "epoch": 0.5, "learning_rate": 9.3622724544909e-06, "loss": 0.283, "step": 1601 }, { "epoch": 0.5, "learning_rate": 9.361872374474897e-06, "loss": 0.2708, "step": 1602 }, { "epoch": 0.5, "learning_rate": 9.361472294458892e-06, "loss": 0.2241, "step": 1603 }, { "epoch": 0.5, "learning_rate": 9.361072214442888e-06, "loss": 0.2674, "step": 1604 }, { "epoch": 0.5, "learning_rate": 9.360672134426885e-06, "loss": 0.308, "step": 1605 }, { "epoch": 0.5, "learning_rate": 9.360272054410883e-06, "loss": 0.2505, "step": 1606 }, { "epoch": 0.5, "learning_rate": 9.35987197439488e-06, "loss": 0.2461, "step": 1607 }, { "epoch": 0.5, "learning_rate": 9.359471894378877e-06, "loss": 0.2405, "step": 1608 }, { "epoch": 0.5, "learning_rate": 9.359071814362873e-06, "loss": 0.3844, "step": 1609 }, { "epoch": 0.5, "learning_rate": 9.35867173434687e-06, "loss": 0.3052, "step": 1610 }, { "epoch": 0.5, "learning_rate": 9.358271654330867e-06, "loss": 0.3425, "step": 1611 }, { "epoch": 0.5, "learning_rate": 9.357871574314864e-06, "loss": 0.2791, "step": 1612 }, { "epoch": 0.5, "learning_rate": 9.35747149429886e-06, "loss": 0.2487, "step": 1613 }, { "epoch": 0.5, "learning_rate": 9.357071414282859e-06, "loss": 0.3284, "step": 1614 }, { "epoch": 0.5, "learning_rate": 9.356671334266854e-06, "loss": 0.2743, "step": 1615 }, { "epoch": 0.5, "learning_rate": 9.35627125425085e-06, "loss": 0.1812, "step": 1616 }, { "epoch": 0.5, "learning_rate": 9.355871174234847e-06, "loss": 0.2897, "step": 1617 }, { "epoch": 0.5, "learning_rate": 9.355471094218844e-06, "loss": 0.37, "step": 1618 }, { "epoch": 0.5, "learning_rate": 9.35507101420284e-06, "loss": 0.364, "step": 1619 }, { "epoch": 0.5, "learning_rate": 9.354670934186839e-06, "loss": 0.3545, "step": 1620 }, { "epoch": 0.5, "learning_rate": 9.354270854170835e-06, "loss": 0.2413, "step": 1621 }, { "epoch": 0.5, "learning_rate": 9.353870774154832e-06, "loss": 0.3222, "step": 1622 }, { "epoch": 0.5, "learning_rate": 9.353470694138829e-06, "loss": 0.181, "step": 1623 }, { "epoch": 0.5, "learning_rate": 9.353070614122825e-06, "loss": 0.2822, "step": 1624 }, { "epoch": 0.5, "learning_rate": 9.352670534106822e-06, "loss": 0.2638, "step": 1625 }, { "epoch": 0.5, "learning_rate": 9.352270454090819e-06, "loss": 0.485, "step": 1626 }, { "epoch": 0.5, "learning_rate": 9.351870374074816e-06, "loss": 0.2666, "step": 1627 }, { "epoch": 0.51, "learning_rate": 9.351470294058812e-06, "loss": 0.1687, "step": 1628 }, { "epoch": 0.51, "learning_rate": 9.351070214042809e-06, "loss": 0.2999, "step": 1629 }, { "epoch": 0.51, "learning_rate": 9.350670134026806e-06, "loss": 0.4002, "step": 1630 }, { "epoch": 0.51, "learning_rate": 9.350270054010802e-06, "loss": 0.2333, "step": 1631 }, { "epoch": 0.51, "learning_rate": 9.349869973994799e-06, "loss": 0.1932, "step": 1632 }, { "epoch": 0.51, "learning_rate": 9.349469893978796e-06, "loss": 0.2386, "step": 1633 }, { "epoch": 0.51, "learning_rate": 9.349069813962794e-06, "loss": 0.2501, "step": 1634 }, { "epoch": 0.51, "learning_rate": 9.34866973394679e-06, "loss": 0.2368, "step": 1635 }, { "epoch": 0.51, "learning_rate": 9.348269653930787e-06, "loss": 0.325, "step": 1636 }, { "epoch": 0.51, "learning_rate": 9.347869573914784e-06, "loss": 0.3124, "step": 1637 }, { "epoch": 0.51, "learning_rate": 9.34746949389878e-06, "loss": 0.4029, "step": 1638 }, { "epoch": 0.51, "learning_rate": 9.347069413882777e-06, "loss": 0.2626, "step": 1639 }, { "epoch": 0.51, "learning_rate": 9.346669333866774e-06, "loss": 0.2689, "step": 1640 }, { "epoch": 0.51, "learning_rate": 9.34626925385077e-06, "loss": 0.3278, "step": 1641 }, { "epoch": 0.51, "learning_rate": 9.345869173834768e-06, "loss": 0.3763, "step": 1642 }, { "epoch": 0.51, "learning_rate": 9.345469093818764e-06, "loss": 0.2583, "step": 1643 }, { "epoch": 0.51, "learning_rate": 9.345069013802761e-06, "loss": 0.2099, "step": 1644 }, { "epoch": 0.51, "learning_rate": 9.344668933786758e-06, "loss": 0.269, "step": 1645 }, { "epoch": 0.51, "learning_rate": 9.344268853770754e-06, "loss": 0.2307, "step": 1646 }, { "epoch": 0.51, "learning_rate": 9.343868773754751e-06, "loss": 0.2287, "step": 1647 }, { "epoch": 0.51, "learning_rate": 9.34346869373875e-06, "loss": 0.3735, "step": 1648 }, { "epoch": 0.51, "learning_rate": 9.343068613722746e-06, "loss": 0.3019, "step": 1649 }, { "epoch": 0.51, "learning_rate": 9.342668533706743e-06, "loss": 0.3393, "step": 1650 }, { "epoch": 0.51, "learning_rate": 9.34226845369074e-06, "loss": 0.3459, "step": 1651 }, { "epoch": 0.51, "learning_rate": 9.341868373674736e-06, "loss": 0.193, "step": 1652 }, { "epoch": 0.51, "learning_rate": 9.341468293658733e-06, "loss": 0.1937, "step": 1653 }, { "epoch": 0.51, "learning_rate": 9.34106821364273e-06, "loss": 0.3375, "step": 1654 }, { "epoch": 0.51, "learning_rate": 9.340668133626726e-06, "loss": 0.2908, "step": 1655 }, { "epoch": 0.51, "learning_rate": 9.340268053610723e-06, "loss": 0.231, "step": 1656 }, { "epoch": 0.51, "learning_rate": 9.33986797359472e-06, "loss": 0.232, "step": 1657 }, { "epoch": 0.51, "learning_rate": 9.339467893578716e-06, "loss": 0.3413, "step": 1658 }, { "epoch": 0.51, "learning_rate": 9.339067813562713e-06, "loss": 0.2114, "step": 1659 }, { "epoch": 0.52, "learning_rate": 9.33866773354671e-06, "loss": 0.3664, "step": 1660 }, { "epoch": 0.52, "learning_rate": 9.338267653530706e-06, "loss": 0.3287, "step": 1661 }, { "epoch": 0.52, "learning_rate": 9.337867573514705e-06, "loss": 0.4126, "step": 1662 }, { "epoch": 0.52, "learning_rate": 9.337467493498701e-06, "loss": 0.1468, "step": 1663 }, { "epoch": 0.52, "learning_rate": 9.337067413482698e-06, "loss": 0.2693, "step": 1664 }, { "epoch": 0.52, "learning_rate": 9.336667333466695e-06, "loss": 0.2212, "step": 1665 }, { "epoch": 0.52, "learning_rate": 9.33626725345069e-06, "loss": 0.3024, "step": 1666 }, { "epoch": 0.52, "learning_rate": 9.335867173434688e-06, "loss": 0.2427, "step": 1667 }, { "epoch": 0.52, "learning_rate": 9.335467093418685e-06, "loss": 0.2462, "step": 1668 }, { "epoch": 0.52, "learning_rate": 9.335067013402681e-06, "loss": 0.3337, "step": 1669 }, { "epoch": 0.52, "learning_rate": 9.334666933386678e-06, "loss": 0.2662, "step": 1670 }, { "epoch": 0.52, "learning_rate": 9.334266853370675e-06, "loss": 0.2713, "step": 1671 }, { "epoch": 0.52, "learning_rate": 9.333866773354672e-06, "loss": 0.2529, "step": 1672 }, { "epoch": 0.52, "learning_rate": 9.333466693338668e-06, "loss": 0.2241, "step": 1673 }, { "epoch": 0.52, "learning_rate": 9.333066613322665e-06, "loss": 0.2889, "step": 1674 }, { "epoch": 0.52, "learning_rate": 9.332666533306662e-06, "loss": 0.3588, "step": 1675 }, { "epoch": 0.52, "learning_rate": 9.33226645329066e-06, "loss": 0.3168, "step": 1676 }, { "epoch": 0.52, "learning_rate": 9.331866373274657e-06, "loss": 0.3266, "step": 1677 }, { "epoch": 0.52, "learning_rate": 9.331466293258652e-06, "loss": 0.2099, "step": 1678 }, { "epoch": 0.52, "learning_rate": 9.331066213242648e-06, "loss": 0.2522, "step": 1679 }, { "epoch": 0.52, "learning_rate": 9.330666133226645e-06, "loss": 0.4058, "step": 1680 }, { "epoch": 0.52, "learning_rate": 9.330266053210643e-06, "loss": 0.2196, "step": 1681 }, { "epoch": 0.52, "learning_rate": 9.32986597319464e-06, "loss": 0.305, "step": 1682 }, { "epoch": 0.52, "learning_rate": 9.329465893178637e-06, "loss": 0.2369, "step": 1683 }, { "epoch": 0.52, "learning_rate": 9.329065813162633e-06, "loss": 0.3609, "step": 1684 }, { "epoch": 0.52, "learning_rate": 9.32866573314663e-06, "loss": 0.2732, "step": 1685 }, { "epoch": 0.52, "learning_rate": 9.328265653130627e-06, "loss": 0.233, "step": 1686 }, { "epoch": 0.52, "learning_rate": 9.327865573114623e-06, "loss": 0.3527, "step": 1687 }, { "epoch": 0.52, "learning_rate": 9.32746549309862e-06, "loss": 0.3302, "step": 1688 }, { "epoch": 0.52, "learning_rate": 9.327065413082617e-06, "loss": 0.3142, "step": 1689 }, { "epoch": 0.52, "learning_rate": 9.326665333066615e-06, "loss": 0.3252, "step": 1690 }, { "epoch": 0.52, "learning_rate": 9.32626525305061e-06, "loss": 0.2212, "step": 1691 }, { "epoch": 0.53, "learning_rate": 9.325865173034607e-06, "loss": 0.4085, "step": 1692 }, { "epoch": 0.53, "learning_rate": 9.325465093018604e-06, "loss": 0.3722, "step": 1693 }, { "epoch": 0.53, "learning_rate": 9.3250650130026e-06, "loss": 0.2747, "step": 1694 }, { "epoch": 0.53, "learning_rate": 9.324664932986599e-06, "loss": 0.341, "step": 1695 }, { "epoch": 0.53, "learning_rate": 9.324264852970595e-06, "loss": 0.2394, "step": 1696 }, { "epoch": 0.53, "learning_rate": 9.323864772954592e-06, "loss": 0.2834, "step": 1697 }, { "epoch": 0.53, "learning_rate": 9.323464692938589e-06, "loss": 0.3301, "step": 1698 }, { "epoch": 0.53, "learning_rate": 9.323064612922585e-06, "loss": 0.2393, "step": 1699 }, { "epoch": 0.53, "learning_rate": 9.322664532906582e-06, "loss": 0.1543, "step": 1700 }, { "epoch": 0.53, "learning_rate": 9.322264452890579e-06, "loss": 0.529, "step": 1701 }, { "epoch": 0.53, "learning_rate": 9.321864372874575e-06, "loss": 0.214, "step": 1702 }, { "epoch": 0.53, "learning_rate": 9.321464292858572e-06, "loss": 0.4842, "step": 1703 }, { "epoch": 0.53, "learning_rate": 9.321064212842569e-06, "loss": 0.3295, "step": 1704 }, { "epoch": 0.53, "learning_rate": 9.320664132826566e-06, "loss": 0.3058, "step": 1705 }, { "epoch": 0.53, "learning_rate": 9.320264052810562e-06, "loss": 0.3592, "step": 1706 }, { "epoch": 0.53, "learning_rate": 9.319863972794559e-06, "loss": 0.2081, "step": 1707 }, { "epoch": 0.53, "learning_rate": 9.319463892778556e-06, "loss": 0.3375, "step": 1708 }, { "epoch": 0.53, "learning_rate": 9.319063812762554e-06, "loss": 0.2311, "step": 1709 }, { "epoch": 0.53, "learning_rate": 9.31866373274655e-06, "loss": 0.2627, "step": 1710 }, { "epoch": 0.53, "learning_rate": 9.318263652730547e-06, "loss": 0.2951, "step": 1711 }, { "epoch": 0.53, "learning_rate": 9.317863572714544e-06, "loss": 0.277, "step": 1712 }, { "epoch": 0.53, "learning_rate": 9.31746349269854e-06, "loss": 0.3108, "step": 1713 }, { "epoch": 0.53, "learning_rate": 9.317063412682537e-06, "loss": 0.1968, "step": 1714 }, { "epoch": 0.53, "learning_rate": 9.316663332666534e-06, "loss": 0.3163, "step": 1715 }, { "epoch": 0.53, "learning_rate": 9.31626325265053e-06, "loss": 0.2904, "step": 1716 }, { "epoch": 0.53, "learning_rate": 9.315863172634527e-06, "loss": 0.276, "step": 1717 }, { "epoch": 0.53, "learning_rate": 9.315463092618524e-06, "loss": 0.3173, "step": 1718 }, { "epoch": 0.53, "learning_rate": 9.31506301260252e-06, "loss": 0.3266, "step": 1719 }, { "epoch": 0.53, "learning_rate": 9.314662932586518e-06, "loss": 0.2227, "step": 1720 }, { "epoch": 0.53, "learning_rate": 9.314262852570514e-06, "loss": 0.1712, "step": 1721 }, { "epoch": 0.53, "learning_rate": 9.313862772554511e-06, "loss": 0.2264, "step": 1722 }, { "epoch": 0.53, "learning_rate": 9.31346269253851e-06, "loss": 0.4462, "step": 1723 }, { "epoch": 0.54, "learning_rate": 9.313062612522506e-06, "loss": 0.1962, "step": 1724 }, { "epoch": 0.54, "learning_rate": 9.312662532506503e-06, "loss": 0.247, "step": 1725 }, { "epoch": 0.54, "learning_rate": 9.3122624524905e-06, "loss": 0.2206, "step": 1726 }, { "epoch": 0.54, "learning_rate": 9.311862372474496e-06, "loss": 0.2545, "step": 1727 }, { "epoch": 0.54, "learning_rate": 9.311462292458493e-06, "loss": 0.1969, "step": 1728 }, { "epoch": 0.54, "learning_rate": 9.31106221244249e-06, "loss": 0.1981, "step": 1729 }, { "epoch": 0.54, "learning_rate": 9.310662132426486e-06, "loss": 0.3012, "step": 1730 }, { "epoch": 0.54, "learning_rate": 9.310262052410483e-06, "loss": 0.2669, "step": 1731 }, { "epoch": 0.54, "learning_rate": 9.30986197239448e-06, "loss": 0.2905, "step": 1732 }, { "epoch": 0.54, "learning_rate": 9.309461892378476e-06, "loss": 0.2641, "step": 1733 }, { "epoch": 0.54, "learning_rate": 9.309061812362473e-06, "loss": 0.288, "step": 1734 }, { "epoch": 0.54, "learning_rate": 9.30866173234647e-06, "loss": 0.1767, "step": 1735 }, { "epoch": 0.54, "learning_rate": 9.308261652330466e-06, "loss": 0.3266, "step": 1736 }, { "epoch": 0.54, "learning_rate": 9.307861572314465e-06, "loss": 0.238, "step": 1737 }, { "epoch": 0.54, "learning_rate": 9.307461492298461e-06, "loss": 0.2152, "step": 1738 }, { "epoch": 0.54, "learning_rate": 9.307061412282458e-06, "loss": 0.3346, "step": 1739 }, { "epoch": 0.54, "learning_rate": 9.306661332266455e-06, "loss": 0.2331, "step": 1740 }, { "epoch": 0.54, "learning_rate": 9.306261252250451e-06, "loss": 0.3262, "step": 1741 }, { "epoch": 0.54, "learning_rate": 9.305861172234448e-06, "loss": 0.2762, "step": 1742 }, { "epoch": 0.54, "learning_rate": 9.305461092218445e-06, "loss": 0.2899, "step": 1743 }, { "epoch": 0.54, "learning_rate": 9.305061012202441e-06, "loss": 0.2884, "step": 1744 }, { "epoch": 0.54, "learning_rate": 9.304660932186438e-06, "loss": 0.5034, "step": 1745 }, { "epoch": 0.54, "learning_rate": 9.304260852170435e-06, "loss": 0.2101, "step": 1746 }, { "epoch": 0.54, "learning_rate": 9.303860772154431e-06, "loss": 0.2543, "step": 1747 }, { "epoch": 0.54, "learning_rate": 9.303460692138428e-06, "loss": 0.2927, "step": 1748 }, { "epoch": 0.54, "learning_rate": 9.303060612122425e-06, "loss": 0.282, "step": 1749 }, { "epoch": 0.54, "learning_rate": 9.302660532106422e-06, "loss": 0.2888, "step": 1750 }, { "epoch": 0.54, "learning_rate": 9.30226045209042e-06, "loss": 0.4351, "step": 1751 }, { "epoch": 0.54, "learning_rate": 9.301860372074417e-06, "loss": 0.2429, "step": 1752 }, { "epoch": 0.54, "learning_rate": 9.301460292058413e-06, "loss": 0.3537, "step": 1753 }, { "epoch": 0.54, "learning_rate": 9.301060212042408e-06, "loss": 0.2323, "step": 1754 }, { "epoch": 0.54, "learning_rate": 9.300660132026405e-06, "loss": 0.2934, "step": 1755 }, { "epoch": 0.55, "learning_rate": 9.300260052010403e-06, "loss": 0.2552, "step": 1756 }, { "epoch": 0.55, "learning_rate": 9.2998599719944e-06, "loss": 0.317, "step": 1757 }, { "epoch": 0.55, "learning_rate": 9.299459891978397e-06, "loss": 0.4095, "step": 1758 }, { "epoch": 0.55, "learning_rate": 9.299059811962393e-06, "loss": 0.3364, "step": 1759 }, { "epoch": 0.55, "learning_rate": 9.29865973194639e-06, "loss": 0.2137, "step": 1760 }, { "epoch": 0.55, "learning_rate": 9.298259651930387e-06, "loss": 0.3346, "step": 1761 }, { "epoch": 0.55, "learning_rate": 9.297859571914383e-06, "loss": 0.3763, "step": 1762 }, { "epoch": 0.55, "learning_rate": 9.29745949189838e-06, "loss": 0.2801, "step": 1763 }, { "epoch": 0.55, "learning_rate": 9.297059411882377e-06, "loss": 0.245, "step": 1764 }, { "epoch": 0.55, "learning_rate": 9.296659331866375e-06, "loss": 0.3369, "step": 1765 }, { "epoch": 0.55, "learning_rate": 9.296259251850372e-06, "loss": 0.2598, "step": 1766 }, { "epoch": 0.55, "learning_rate": 9.295859171834367e-06, "loss": 0.3596, "step": 1767 }, { "epoch": 0.55, "learning_rate": 9.295459091818364e-06, "loss": 0.3702, "step": 1768 }, { "epoch": 0.55, "learning_rate": 9.29505901180236e-06, "loss": 0.323, "step": 1769 }, { "epoch": 0.55, "learning_rate": 9.294658931786359e-06, "loss": 0.4557, "step": 1770 }, { "epoch": 0.55, "learning_rate": 9.294258851770355e-06, "loss": 0.4469, "step": 1771 }, { "epoch": 0.55, "learning_rate": 9.293858771754352e-06, "loss": 0.2855, "step": 1772 }, { "epoch": 0.55, "learning_rate": 9.293458691738349e-06, "loss": 0.3459, "step": 1773 }, { "epoch": 0.55, "learning_rate": 9.293058611722345e-06, "loss": 0.2368, "step": 1774 }, { "epoch": 0.55, "learning_rate": 9.292658531706342e-06, "loss": 0.2854, "step": 1775 }, { "epoch": 0.55, "learning_rate": 9.292258451690339e-06, "loss": 0.3901, "step": 1776 }, { "epoch": 0.55, "learning_rate": 9.291858371674335e-06, "loss": 0.154, "step": 1777 }, { "epoch": 0.55, "learning_rate": 9.291458291658332e-06, "loss": 0.2676, "step": 1778 }, { "epoch": 0.55, "learning_rate": 9.291058211642329e-06, "loss": 0.2812, "step": 1779 }, { "epoch": 0.55, "learning_rate": 9.290658131626325e-06, "loss": 0.2833, "step": 1780 }, { "epoch": 0.55, "learning_rate": 9.290258051610322e-06, "loss": 0.2826, "step": 1781 }, { "epoch": 0.55, "learning_rate": 9.289857971594319e-06, "loss": 0.2525, "step": 1782 }, { "epoch": 0.55, "learning_rate": 9.289457891578316e-06, "loss": 0.4154, "step": 1783 }, { "epoch": 0.55, "learning_rate": 9.289057811562314e-06, "loss": 0.2469, "step": 1784 }, { "epoch": 0.55, "learning_rate": 9.28865773154631e-06, "loss": 0.2629, "step": 1785 }, { "epoch": 0.55, "learning_rate": 9.288257651530307e-06, "loss": 0.2177, "step": 1786 }, { "epoch": 0.55, "learning_rate": 9.287857571514304e-06, "loss": 0.2379, "step": 1787 }, { "epoch": 0.55, "learning_rate": 9.2874574914983e-06, "loss": 0.2967, "step": 1788 }, { "epoch": 0.56, "learning_rate": 9.287057411482297e-06, "loss": 0.3889, "step": 1789 }, { "epoch": 0.56, "learning_rate": 9.286657331466294e-06, "loss": 0.2506, "step": 1790 }, { "epoch": 0.56, "learning_rate": 9.28625725145029e-06, "loss": 0.3745, "step": 1791 }, { "epoch": 0.56, "learning_rate": 9.285857171434287e-06, "loss": 0.2312, "step": 1792 }, { "epoch": 0.56, "learning_rate": 9.285457091418284e-06, "loss": 0.2538, "step": 1793 }, { "epoch": 0.56, "learning_rate": 9.28505701140228e-06, "loss": 0.3809, "step": 1794 }, { "epoch": 0.56, "learning_rate": 9.284656931386277e-06, "loss": 0.1778, "step": 1795 }, { "epoch": 0.56, "learning_rate": 9.284256851370274e-06, "loss": 0.197, "step": 1796 }, { "epoch": 0.56, "learning_rate": 9.28385677135427e-06, "loss": 0.3561, "step": 1797 }, { "epoch": 0.56, "learning_rate": 9.28345669133827e-06, "loss": 0.2072, "step": 1798 }, { "epoch": 0.56, "learning_rate": 9.283056611322266e-06, "loss": 0.2615, "step": 1799 }, { "epoch": 0.56, "learning_rate": 9.282656531306263e-06, "loss": 0.432, "step": 1800 }, { "epoch": 0.56, "learning_rate": 9.28225645129026e-06, "loss": 0.2521, "step": 1801 }, { "epoch": 0.56, "learning_rate": 9.281856371274256e-06, "loss": 0.2315, "step": 1802 }, { "epoch": 0.56, "learning_rate": 9.281456291258253e-06, "loss": 0.4432, "step": 1803 }, { "epoch": 0.56, "learning_rate": 9.28105621124225e-06, "loss": 0.2514, "step": 1804 }, { "epoch": 0.56, "learning_rate": 9.280656131226246e-06, "loss": 0.4187, "step": 1805 }, { "epoch": 0.56, "learning_rate": 9.280256051210243e-06, "loss": 0.2941, "step": 1806 }, { "epoch": 0.56, "learning_rate": 9.27985597119424e-06, "loss": 0.2346, "step": 1807 }, { "epoch": 0.56, "learning_rate": 9.279455891178236e-06, "loss": 0.3942, "step": 1808 }, { "epoch": 0.56, "learning_rate": 9.279055811162233e-06, "loss": 0.2731, "step": 1809 }, { "epoch": 0.56, "learning_rate": 9.27865573114623e-06, "loss": 0.3871, "step": 1810 }, { "epoch": 0.56, "learning_rate": 9.278255651130226e-06, "loss": 0.2717, "step": 1811 }, { "epoch": 0.56, "learning_rate": 9.277855571114225e-06, "loss": 0.2729, "step": 1812 }, { "epoch": 0.56, "learning_rate": 9.277455491098221e-06, "loss": 0.4008, "step": 1813 }, { "epoch": 0.56, "learning_rate": 9.277055411082218e-06, "loss": 0.2503, "step": 1814 }, { "epoch": 0.56, "learning_rate": 9.276655331066215e-06, "loss": 0.3559, "step": 1815 }, { "epoch": 0.56, "learning_rate": 9.276255251050211e-06, "loss": 0.2439, "step": 1816 }, { "epoch": 0.56, "learning_rate": 9.275855171034206e-06, "loss": 0.2714, "step": 1817 }, { "epoch": 0.56, "learning_rate": 9.275455091018205e-06, "loss": 0.2517, "step": 1818 }, { "epoch": 0.56, "learning_rate": 9.275055011002201e-06, "loss": 0.2379, "step": 1819 }, { "epoch": 0.56, "learning_rate": 9.274654930986198e-06, "loss": 0.1971, "step": 1820 }, { "epoch": 0.57, "learning_rate": 9.274254850970195e-06, "loss": 0.1736, "step": 1821 }, { "epoch": 0.57, "learning_rate": 9.273854770954191e-06, "loss": 0.3974, "step": 1822 }, { "epoch": 0.57, "learning_rate": 9.273454690938188e-06, "loss": 0.1887, "step": 1823 }, { "epoch": 0.57, "learning_rate": 9.273054610922185e-06, "loss": 0.2116, "step": 1824 }, { "epoch": 0.57, "learning_rate": 9.272654530906181e-06, "loss": 0.1796, "step": 1825 }, { "epoch": 0.57, "learning_rate": 9.27225445089018e-06, "loss": 0.2869, "step": 1826 }, { "epoch": 0.57, "learning_rate": 9.271854370874177e-06, "loss": 0.2707, "step": 1827 }, { "epoch": 0.57, "learning_rate": 9.271454290858173e-06, "loss": 0.2958, "step": 1828 }, { "epoch": 0.57, "learning_rate": 9.27105421084217e-06, "loss": 0.3007, "step": 1829 }, { "epoch": 0.57, "learning_rate": 9.270654130826165e-06, "loss": 0.271, "step": 1830 }, { "epoch": 0.57, "learning_rate": 9.270254050810162e-06, "loss": 0.381, "step": 1831 }, { "epoch": 0.57, "learning_rate": 9.26985397079416e-06, "loss": 0.2709, "step": 1832 }, { "epoch": 0.57, "learning_rate": 9.269453890778157e-06, "loss": 0.3045, "step": 1833 }, { "epoch": 0.57, "learning_rate": 9.269053810762153e-06, "loss": 0.2614, "step": 1834 }, { "epoch": 0.57, "learning_rate": 9.26865373074615e-06, "loss": 0.276, "step": 1835 }, { "epoch": 0.57, "learning_rate": 9.268253650730147e-06, "loss": 0.2947, "step": 1836 }, { "epoch": 0.57, "learning_rate": 9.267853570714143e-06, "loss": 0.2643, "step": 1837 }, { "epoch": 0.57, "learning_rate": 9.26745349069814e-06, "loss": 0.2863, "step": 1838 }, { "epoch": 0.57, "learning_rate": 9.267053410682137e-06, "loss": 0.2541, "step": 1839 }, { "epoch": 0.57, "learning_rate": 9.266653330666135e-06, "loss": 0.4018, "step": 1840 }, { "epoch": 0.57, "learning_rate": 9.266253250650132e-06, "loss": 0.3455, "step": 1841 }, { "epoch": 0.57, "learning_rate": 9.265853170634127e-06, "loss": 0.1762, "step": 1842 }, { "epoch": 0.57, "learning_rate": 9.265453090618123e-06, "loss": 0.3026, "step": 1843 }, { "epoch": 0.57, "learning_rate": 9.26505301060212e-06, "loss": 0.2965, "step": 1844 }, { "epoch": 0.57, "learning_rate": 9.264652930586117e-06, "loss": 0.3598, "step": 1845 }, { "epoch": 0.57, "learning_rate": 9.264252850570115e-06, "loss": 0.2465, "step": 1846 }, { "epoch": 0.57, "learning_rate": 9.263852770554112e-06, "loss": 0.2573, "step": 1847 }, { "epoch": 0.57, "learning_rate": 9.263452690538109e-06, "loss": 0.2794, "step": 1848 }, { "epoch": 0.57, "learning_rate": 9.263052610522105e-06, "loss": 0.3703, "step": 1849 }, { "epoch": 0.57, "learning_rate": 9.262652530506102e-06, "loss": 0.2255, "step": 1850 }, { "epoch": 0.57, "learning_rate": 9.262252450490099e-06, "loss": 0.5259, "step": 1851 }, { "epoch": 0.57, "learning_rate": 9.261852370474095e-06, "loss": 0.2293, "step": 1852 }, { "epoch": 0.58, "learning_rate": 9.261452290458092e-06, "loss": 0.1983, "step": 1853 }, { "epoch": 0.58, "learning_rate": 9.26105221044209e-06, "loss": 0.2519, "step": 1854 }, { "epoch": 0.58, "learning_rate": 9.260652130426085e-06, "loss": 0.3376, "step": 1855 }, { "epoch": 0.58, "learning_rate": 9.260252050410082e-06, "loss": 0.4481, "step": 1856 }, { "epoch": 0.58, "learning_rate": 9.259851970394079e-06, "loss": 0.3842, "step": 1857 }, { "epoch": 0.58, "learning_rate": 9.259451890378075e-06, "loss": 0.5303, "step": 1858 }, { "epoch": 0.58, "learning_rate": 9.259051810362072e-06, "loss": 0.2716, "step": 1859 }, { "epoch": 0.58, "learning_rate": 9.25865173034607e-06, "loss": 0.3253, "step": 1860 }, { "epoch": 0.58, "learning_rate": 9.258251650330067e-06, "loss": 0.241, "step": 1861 }, { "epoch": 0.58, "learning_rate": 9.257851570314064e-06, "loss": 0.4547, "step": 1862 }, { "epoch": 0.58, "learning_rate": 9.25745149029806e-06, "loss": 0.2204, "step": 1863 }, { "epoch": 0.58, "learning_rate": 9.257051410282057e-06, "loss": 0.2375, "step": 1864 }, { "epoch": 0.58, "learning_rate": 9.256651330266054e-06, "loss": 0.2617, "step": 1865 }, { "epoch": 0.58, "learning_rate": 9.25625125025005e-06, "loss": 0.3414, "step": 1866 }, { "epoch": 0.58, "learning_rate": 9.255851170234047e-06, "loss": 0.1908, "step": 1867 }, { "epoch": 0.58, "learning_rate": 9.255451090218044e-06, "loss": 0.3207, "step": 1868 }, { "epoch": 0.58, "learning_rate": 9.25505101020204e-06, "loss": 0.2584, "step": 1869 }, { "epoch": 0.58, "learning_rate": 9.254650930186037e-06, "loss": 0.287, "step": 1870 }, { "epoch": 0.58, "learning_rate": 9.254250850170034e-06, "loss": 0.3343, "step": 1871 }, { "epoch": 0.58, "learning_rate": 9.25385077015403e-06, "loss": 0.2982, "step": 1872 }, { "epoch": 0.58, "learning_rate": 9.25345069013803e-06, "loss": 0.3308, "step": 1873 }, { "epoch": 0.58, "learning_rate": 9.253050610122026e-06, "loss": 0.298, "step": 1874 }, { "epoch": 0.58, "learning_rate": 9.252650530106023e-06, "loss": 0.2713, "step": 1875 }, { "epoch": 0.58, "learning_rate": 9.25225045009002e-06, "loss": 0.2508, "step": 1876 }, { "epoch": 0.58, "learning_rate": 9.251850370074016e-06, "loss": 0.248, "step": 1877 }, { "epoch": 0.58, "learning_rate": 9.251450290058013e-06, "loss": 0.2159, "step": 1878 }, { "epoch": 0.58, "learning_rate": 9.25105021004201e-06, "loss": 0.2907, "step": 1879 }, { "epoch": 0.58, "learning_rate": 9.250650130026006e-06, "loss": 0.2531, "step": 1880 }, { "epoch": 0.58, "learning_rate": 9.250250050010003e-06, "loss": 0.277, "step": 1881 }, { "epoch": 0.58, "learning_rate": 9.249849969994e-06, "loss": 0.4171, "step": 1882 }, { "epoch": 0.58, "learning_rate": 9.249449889977996e-06, "loss": 0.2723, "step": 1883 }, { "epoch": 0.58, "learning_rate": 9.249049809961993e-06, "loss": 0.2413, "step": 1884 }, { "epoch": 0.59, "learning_rate": 9.24864972994599e-06, "loss": 0.1917, "step": 1885 }, { "epoch": 0.59, "learning_rate": 9.248249649929986e-06, "loss": 0.1858, "step": 1886 }, { "epoch": 0.59, "learning_rate": 9.247849569913984e-06, "loss": 0.2699, "step": 1887 }, { "epoch": 0.59, "learning_rate": 9.247449489897981e-06, "loss": 0.2933, "step": 1888 }, { "epoch": 0.59, "learning_rate": 9.247049409881978e-06, "loss": 0.3357, "step": 1889 }, { "epoch": 0.59, "learning_rate": 9.246649329865975e-06, "loss": 0.2489, "step": 1890 }, { "epoch": 0.59, "learning_rate": 9.246249249849971e-06, "loss": 0.3504, "step": 1891 }, { "epoch": 0.59, "learning_rate": 9.245849169833968e-06, "loss": 0.459, "step": 1892 }, { "epoch": 0.59, "learning_rate": 9.245449089817965e-06, "loss": 0.1906, "step": 1893 }, { "epoch": 0.59, "learning_rate": 9.245049009801961e-06, "loss": 0.4935, "step": 1894 }, { "epoch": 0.59, "learning_rate": 9.244648929785958e-06, "loss": 0.2443, "step": 1895 }, { "epoch": 0.59, "learning_rate": 9.244248849769955e-06, "loss": 0.2703, "step": 1896 }, { "epoch": 0.59, "learning_rate": 9.243848769753951e-06, "loss": 0.339, "step": 1897 }, { "epoch": 0.59, "learning_rate": 9.243448689737948e-06, "loss": 0.1523, "step": 1898 }, { "epoch": 0.59, "learning_rate": 9.243048609721945e-06, "loss": 0.3796, "step": 1899 }, { "epoch": 0.59, "learning_rate": 9.242648529705941e-06, "loss": 0.2438, "step": 1900 }, { "epoch": 0.59, "learning_rate": 9.24224844968994e-06, "loss": 0.3532, "step": 1901 }, { "epoch": 0.59, "learning_rate": 9.241848369673936e-06, "loss": 0.322, "step": 1902 }, { "epoch": 0.59, "learning_rate": 9.241448289657933e-06, "loss": 0.2499, "step": 1903 }, { "epoch": 0.59, "learning_rate": 9.24104820964193e-06, "loss": 0.256, "step": 1904 }, { "epoch": 0.59, "learning_rate": 9.240648129625925e-06, "loss": 0.2556, "step": 1905 }, { "epoch": 0.59, "learning_rate": 9.240248049609922e-06, "loss": 0.2776, "step": 1906 }, { "epoch": 0.59, "learning_rate": 9.23984796959392e-06, "loss": 0.2382, "step": 1907 }, { "epoch": 0.59, "learning_rate": 9.239447889577917e-06, "loss": 0.203, "step": 1908 }, { "epoch": 0.59, "learning_rate": 9.239047809561913e-06, "loss": 0.3452, "step": 1909 }, { "epoch": 0.59, "learning_rate": 9.23864772954591e-06, "loss": 0.2717, "step": 1910 }, { "epoch": 0.59, "learning_rate": 9.238247649529907e-06, "loss": 0.3764, "step": 1911 }, { "epoch": 0.59, "learning_rate": 9.237847569513903e-06, "loss": 0.2006, "step": 1912 }, { "epoch": 0.59, "learning_rate": 9.2374474894979e-06, "loss": 0.2449, "step": 1913 }, { "epoch": 0.59, "learning_rate": 9.237047409481897e-06, "loss": 0.1398, "step": 1914 }, { "epoch": 0.59, "learning_rate": 9.236647329465895e-06, "loss": 0.3136, "step": 1915 }, { "epoch": 0.59, "learning_rate": 9.236247249449892e-06, "loss": 0.2151, "step": 1916 }, { "epoch": 0.59, "learning_rate": 9.235847169433888e-06, "loss": 0.267, "step": 1917 }, { "epoch": 0.6, "learning_rate": 9.235447089417883e-06, "loss": 0.2552, "step": 1918 }, { "epoch": 0.6, "learning_rate": 9.23504700940188e-06, "loss": 0.3741, "step": 1919 }, { "epoch": 0.6, "learning_rate": 9.234646929385877e-06, "loss": 0.14, "step": 1920 }, { "epoch": 0.6, "learning_rate": 9.234246849369875e-06, "loss": 0.2829, "step": 1921 }, { "epoch": 0.6, "learning_rate": 9.233846769353872e-06, "loss": 0.2428, "step": 1922 }, { "epoch": 0.6, "learning_rate": 9.233446689337869e-06, "loss": 0.2128, "step": 1923 }, { "epoch": 0.6, "learning_rate": 9.233046609321865e-06, "loss": 0.1992, "step": 1924 }, { "epoch": 0.6, "learning_rate": 9.232646529305862e-06, "loss": 0.2944, "step": 1925 }, { "epoch": 0.6, "learning_rate": 9.232246449289859e-06, "loss": 0.2107, "step": 1926 }, { "epoch": 0.6, "learning_rate": 9.231846369273855e-06, "loss": 0.2178, "step": 1927 }, { "epoch": 0.6, "learning_rate": 9.231446289257852e-06, "loss": 0.3941, "step": 1928 }, { "epoch": 0.6, "learning_rate": 9.23104620924185e-06, "loss": 0.2458, "step": 1929 }, { "epoch": 0.6, "learning_rate": 9.230646129225847e-06, "loss": 0.2383, "step": 1930 }, { "epoch": 0.6, "learning_rate": 9.230246049209842e-06, "loss": 0.2853, "step": 1931 }, { "epoch": 0.6, "learning_rate": 9.229845969193839e-06, "loss": 0.3656, "step": 1932 }, { "epoch": 0.6, "learning_rate": 9.229445889177835e-06, "loss": 0.3998, "step": 1933 }, { "epoch": 0.6, "learning_rate": 9.229045809161832e-06, "loss": 0.2838, "step": 1934 }, { "epoch": 0.6, "learning_rate": 9.22864572914583e-06, "loss": 0.3461, "step": 1935 }, { "epoch": 0.6, "learning_rate": 9.228245649129827e-06, "loss": 0.3325, "step": 1936 }, { "epoch": 0.6, "learning_rate": 9.227845569113824e-06, "loss": 0.4186, "step": 1937 }, { "epoch": 0.6, "learning_rate": 9.22744548909782e-06, "loss": 0.1804, "step": 1938 }, { "epoch": 0.6, "learning_rate": 9.227045409081817e-06, "loss": 0.2483, "step": 1939 }, { "epoch": 0.6, "learning_rate": 9.226645329065814e-06, "loss": 0.2438, "step": 1940 }, { "epoch": 0.6, "learning_rate": 9.22624524904981e-06, "loss": 0.1944, "step": 1941 }, { "epoch": 0.6, "learning_rate": 9.225845169033807e-06, "loss": 0.2435, "step": 1942 }, { "epoch": 0.6, "learning_rate": 9.225445089017804e-06, "loss": 0.2962, "step": 1943 }, { "epoch": 0.6, "learning_rate": 9.2250450090018e-06, "loss": 0.1609, "step": 1944 }, { "epoch": 0.6, "learning_rate": 9.224644928985797e-06, "loss": 0.4199, "step": 1945 }, { "epoch": 0.6, "learning_rate": 9.224244848969794e-06, "loss": 0.3177, "step": 1946 }, { "epoch": 0.6, "learning_rate": 9.22384476895379e-06, "loss": 0.2881, "step": 1947 }, { "epoch": 0.6, "learning_rate": 9.223444688937787e-06, "loss": 0.2052, "step": 1948 }, { "epoch": 0.6, "learning_rate": 9.223044608921786e-06, "loss": 0.3221, "step": 1949 }, { "epoch": 0.61, "learning_rate": 9.222644528905782e-06, "loss": 0.2313, "step": 1950 }, { "epoch": 0.61, "learning_rate": 9.22224444888978e-06, "loss": 0.1771, "step": 1951 }, { "epoch": 0.61, "learning_rate": 9.221844368873776e-06, "loss": 0.179, "step": 1952 }, { "epoch": 0.61, "learning_rate": 9.221444288857773e-06, "loss": 0.2934, "step": 1953 }, { "epoch": 0.61, "learning_rate": 9.22104420884177e-06, "loss": 0.309, "step": 1954 }, { "epoch": 0.61, "learning_rate": 9.220644128825766e-06, "loss": 0.325, "step": 1955 }, { "epoch": 0.61, "learning_rate": 9.220244048809763e-06, "loss": 0.2088, "step": 1956 }, { "epoch": 0.61, "learning_rate": 9.21984396879376e-06, "loss": 0.2736, "step": 1957 }, { "epoch": 0.61, "learning_rate": 9.219443888777756e-06, "loss": 0.259, "step": 1958 }, { "epoch": 0.61, "learning_rate": 9.219043808761753e-06, "loss": 0.2985, "step": 1959 }, { "epoch": 0.61, "learning_rate": 9.21864372874575e-06, "loss": 0.3288, "step": 1960 }, { "epoch": 0.61, "learning_rate": 9.218243648729746e-06, "loss": 0.2273, "step": 1961 }, { "epoch": 0.61, "learning_rate": 9.217843568713743e-06, "loss": 0.2875, "step": 1962 }, { "epoch": 0.61, "learning_rate": 9.217443488697741e-06, "loss": 0.3496, "step": 1963 }, { "epoch": 0.61, "learning_rate": 9.217043408681738e-06, "loss": 0.4037, "step": 1964 }, { "epoch": 0.61, "learning_rate": 9.216643328665734e-06, "loss": 0.2213, "step": 1965 }, { "epoch": 0.61, "learning_rate": 9.216243248649731e-06, "loss": 0.3121, "step": 1966 }, { "epoch": 0.61, "learning_rate": 9.215843168633728e-06, "loss": 0.2237, "step": 1967 }, { "epoch": 0.61, "learning_rate": 9.215443088617725e-06, "loss": 0.2868, "step": 1968 }, { "epoch": 0.61, "learning_rate": 9.215043008601721e-06, "loss": 0.337, "step": 1969 }, { "epoch": 0.61, "learning_rate": 9.214642928585718e-06, "loss": 0.3803, "step": 1970 }, { "epoch": 0.61, "learning_rate": 9.214242848569715e-06, "loss": 0.2534, "step": 1971 }, { "epoch": 0.61, "learning_rate": 9.213842768553711e-06, "loss": 0.2764, "step": 1972 }, { "epoch": 0.61, "learning_rate": 9.213442688537708e-06, "loss": 0.1913, "step": 1973 }, { "epoch": 0.61, "learning_rate": 9.213042608521705e-06, "loss": 0.3404, "step": 1974 }, { "epoch": 0.61, "learning_rate": 9.212642528505701e-06, "loss": 0.3313, "step": 1975 }, { "epoch": 0.61, "learning_rate": 9.212242448489698e-06, "loss": 0.3746, "step": 1976 }, { "epoch": 0.61, "learning_rate": 9.211842368473696e-06, "loss": 0.3171, "step": 1977 }, { "epoch": 0.61, "learning_rate": 9.211442288457693e-06, "loss": 0.2355, "step": 1978 }, { "epoch": 0.61, "learning_rate": 9.21104220844169e-06, "loss": 0.2645, "step": 1979 }, { "epoch": 0.61, "learning_rate": 9.210642128425686e-06, "loss": 0.2411, "step": 1980 }, { "epoch": 0.61, "learning_rate": 9.210242048409681e-06, "loss": 0.3396, "step": 1981 }, { "epoch": 0.62, "learning_rate": 9.20984196839368e-06, "loss": 0.3318, "step": 1982 }, { "epoch": 0.62, "learning_rate": 9.209441888377677e-06, "loss": 0.3958, "step": 1983 }, { "epoch": 0.62, "learning_rate": 9.209041808361673e-06, "loss": 0.424, "step": 1984 }, { "epoch": 0.62, "learning_rate": 9.20864172834567e-06, "loss": 0.4421, "step": 1985 }, { "epoch": 0.62, "learning_rate": 9.208241648329667e-06, "loss": 0.468, "step": 1986 }, { "epoch": 0.62, "learning_rate": 9.207841568313663e-06, "loss": 0.3915, "step": 1987 }, { "epoch": 0.62, "learning_rate": 9.20744148829766e-06, "loss": 0.2198, "step": 1988 }, { "epoch": 0.62, "learning_rate": 9.207041408281657e-06, "loss": 0.1696, "step": 1989 }, { "epoch": 0.62, "learning_rate": 9.206641328265653e-06, "loss": 0.3092, "step": 1990 }, { "epoch": 0.62, "learning_rate": 9.206241248249652e-06, "loss": 0.3053, "step": 1991 }, { "epoch": 0.62, "learning_rate": 9.205841168233648e-06, "loss": 0.1993, "step": 1992 }, { "epoch": 0.62, "learning_rate": 9.205441088217645e-06, "loss": 0.3543, "step": 1993 }, { "epoch": 0.62, "learning_rate": 9.20504100820164e-06, "loss": 0.2286, "step": 1994 }, { "epoch": 0.62, "learning_rate": 9.204640928185637e-06, "loss": 0.2934, "step": 1995 }, { "epoch": 0.62, "learning_rate": 9.204240848169635e-06, "loss": 0.3263, "step": 1996 }, { "epoch": 0.62, "learning_rate": 9.203840768153632e-06, "loss": 0.4326, "step": 1997 }, { "epoch": 0.62, "learning_rate": 9.203440688137629e-06, "loss": 0.2941, "step": 1998 }, { "epoch": 0.62, "learning_rate": 9.203040608121625e-06, "loss": 0.202, "step": 1999 }, { "epoch": 0.62, "learning_rate": 9.202640528105622e-06, "loss": 0.1899, "step": 2000 }, { "epoch": 0.62, "learning_rate": 9.202240448089619e-06, "loss": 0.2474, "step": 2001 }, { "epoch": 0.62, "learning_rate": 9.201840368073615e-06, "loss": 0.2045, "step": 2002 }, { "epoch": 0.62, "learning_rate": 9.201440288057612e-06, "loss": 0.3908, "step": 2003 }, { "epoch": 0.62, "learning_rate": 9.201040208041609e-06, "loss": 0.3965, "step": 2004 }, { "epoch": 0.62, "learning_rate": 9.200640128025607e-06, "loss": 0.3248, "step": 2005 }, { "epoch": 0.62, "learning_rate": 9.200240048009602e-06, "loss": 0.4128, "step": 2006 }, { "epoch": 0.62, "learning_rate": 9.199839967993599e-06, "loss": 0.2846, "step": 2007 }, { "epoch": 0.62, "learning_rate": 9.199439887977595e-06, "loss": 0.2482, "step": 2008 }, { "epoch": 0.62, "learning_rate": 9.199039807961592e-06, "loss": 0.2765, "step": 2009 }, { "epoch": 0.62, "learning_rate": 9.19863972794559e-06, "loss": 0.2036, "step": 2010 }, { "epoch": 0.62, "learning_rate": 9.198239647929587e-06, "loss": 0.2455, "step": 2011 }, { "epoch": 0.62, "learning_rate": 9.197839567913584e-06, "loss": 0.2607, "step": 2012 }, { "epoch": 0.62, "learning_rate": 9.19743948789758e-06, "loss": 0.2383, "step": 2013 }, { "epoch": 0.63, "learning_rate": 9.197039407881577e-06, "loss": 0.1731, "step": 2014 }, { "epoch": 0.63, "learning_rate": 9.196639327865574e-06, "loss": 0.267, "step": 2015 }, { "epoch": 0.63, "learning_rate": 9.19623924784957e-06, "loss": 0.2848, "step": 2016 }, { "epoch": 0.63, "learning_rate": 9.195839167833567e-06, "loss": 0.4632, "step": 2017 }, { "epoch": 0.63, "learning_rate": 9.195439087817566e-06, "loss": 0.4539, "step": 2018 }, { "epoch": 0.63, "learning_rate": 9.19503900780156e-06, "loss": 0.2481, "step": 2019 }, { "epoch": 0.63, "learning_rate": 9.194638927785557e-06, "loss": 0.2174, "step": 2020 }, { "epoch": 0.63, "learning_rate": 9.194238847769554e-06, "loss": 0.4125, "step": 2021 }, { "epoch": 0.63, "learning_rate": 9.19383876775355e-06, "loss": 0.1576, "step": 2022 }, { "epoch": 0.63, "learning_rate": 9.193438687737547e-06, "loss": 0.3914, "step": 2023 }, { "epoch": 0.63, "learning_rate": 9.193038607721546e-06, "loss": 0.2226, "step": 2024 }, { "epoch": 0.63, "learning_rate": 9.192638527705542e-06, "loss": 0.2506, "step": 2025 }, { "epoch": 0.63, "learning_rate": 9.192238447689539e-06, "loss": 0.2697, "step": 2026 }, { "epoch": 0.63, "learning_rate": 9.191838367673536e-06, "loss": 0.2858, "step": 2027 }, { "epoch": 0.63, "learning_rate": 9.191438287657532e-06, "loss": 0.1233, "step": 2028 }, { "epoch": 0.63, "learning_rate": 9.19103820764153e-06, "loss": 0.3177, "step": 2029 }, { "epoch": 0.63, "learning_rate": 9.190638127625526e-06, "loss": 0.1463, "step": 2030 }, { "epoch": 0.63, "learning_rate": 9.190238047609523e-06, "loss": 0.2456, "step": 2031 }, { "epoch": 0.63, "learning_rate": 9.18983796759352e-06, "loss": 0.1233, "step": 2032 }, { "epoch": 0.63, "learning_rate": 9.189437887577516e-06, "loss": 0.3316, "step": 2033 }, { "epoch": 0.63, "learning_rate": 9.189037807561513e-06, "loss": 0.1765, "step": 2034 }, { "epoch": 0.63, "learning_rate": 9.18863772754551e-06, "loss": 0.2992, "step": 2035 }, { "epoch": 0.63, "learning_rate": 9.188237647529506e-06, "loss": 0.4301, "step": 2036 }, { "epoch": 0.63, "learning_rate": 9.187837567513503e-06, "loss": 0.2693, "step": 2037 }, { "epoch": 0.63, "learning_rate": 9.187437487497501e-06, "loss": 0.2254, "step": 2038 }, { "epoch": 0.63, "learning_rate": 9.187037407481498e-06, "loss": 0.1987, "step": 2039 }, { "epoch": 0.63, "learning_rate": 9.186637327465494e-06, "loss": 0.4002, "step": 2040 }, { "epoch": 0.63, "learning_rate": 9.186237247449491e-06, "loss": 0.276, "step": 2041 }, { "epoch": 0.63, "learning_rate": 9.185837167433488e-06, "loss": 0.346, "step": 2042 }, { "epoch": 0.63, "learning_rate": 9.185437087417484e-06, "loss": 0.3554, "step": 2043 }, { "epoch": 0.63, "learning_rate": 9.185037007401481e-06, "loss": 0.2239, "step": 2044 }, { "epoch": 0.63, "learning_rate": 9.184636927385478e-06, "loss": 0.2902, "step": 2045 }, { "epoch": 0.64, "learning_rate": 9.184236847369475e-06, "loss": 0.269, "step": 2046 }, { "epoch": 0.64, "learning_rate": 9.183836767353471e-06, "loss": 0.2687, "step": 2047 }, { "epoch": 0.64, "learning_rate": 9.183436687337468e-06, "loss": 0.311, "step": 2048 }, { "epoch": 0.64, "learning_rate": 9.183036607321465e-06, "loss": 0.1727, "step": 2049 }, { "epoch": 0.64, "learning_rate": 9.182636527305461e-06, "loss": 0.2992, "step": 2050 }, { "epoch": 0.64, "learning_rate": 9.182236447289458e-06, "loss": 0.4523, "step": 2051 }, { "epoch": 0.64, "learning_rate": 9.181836367273456e-06, "loss": 0.3023, "step": 2052 }, { "epoch": 0.64, "learning_rate": 9.181436287257453e-06, "loss": 0.3871, "step": 2053 }, { "epoch": 0.64, "learning_rate": 9.18103620724145e-06, "loss": 0.5112, "step": 2054 }, { "epoch": 0.64, "learning_rate": 9.180636127225446e-06, "loss": 0.2225, "step": 2055 }, { "epoch": 0.64, "learning_rate": 9.180236047209443e-06, "loss": 0.2299, "step": 2056 }, { "epoch": 0.64, "learning_rate": 9.17983596719344e-06, "loss": 0.448, "step": 2057 }, { "epoch": 0.64, "learning_rate": 9.179435887177436e-06, "loss": 0.1927, "step": 2058 }, { "epoch": 0.64, "learning_rate": 9.179035807161433e-06, "loss": 0.356, "step": 2059 }, { "epoch": 0.64, "learning_rate": 9.17863572714543e-06, "loss": 0.2521, "step": 2060 }, { "epoch": 0.64, "learning_rate": 9.178235647129427e-06, "loss": 0.2601, "step": 2061 }, { "epoch": 0.64, "learning_rate": 9.177835567113423e-06, "loss": 0.2293, "step": 2062 }, { "epoch": 0.64, "learning_rate": 9.17743548709742e-06, "loss": 0.33, "step": 2063 }, { "epoch": 0.64, "learning_rate": 9.177035407081417e-06, "loss": 0.1865, "step": 2064 }, { "epoch": 0.64, "learning_rate": 9.176635327065413e-06, "loss": 0.3684, "step": 2065 }, { "epoch": 0.64, "learning_rate": 9.176235247049412e-06, "loss": 0.3519, "step": 2066 }, { "epoch": 0.64, "learning_rate": 9.175835167033408e-06, "loss": 0.2484, "step": 2067 }, { "epoch": 0.64, "learning_rate": 9.175435087017405e-06, "loss": 0.343, "step": 2068 }, { "epoch": 0.64, "learning_rate": 9.1750350070014e-06, "loss": 0.2923, "step": 2069 }, { "epoch": 0.64, "learning_rate": 9.174634926985397e-06, "loss": 0.2826, "step": 2070 }, { "epoch": 0.64, "learning_rate": 9.174234846969395e-06, "loss": 0.3379, "step": 2071 }, { "epoch": 0.64, "learning_rate": 9.173834766953392e-06, "loss": 0.2169, "step": 2072 }, { "epoch": 0.64, "learning_rate": 9.173434686937388e-06, "loss": 0.2529, "step": 2073 }, { "epoch": 0.64, "learning_rate": 9.173034606921385e-06, "loss": 0.4357, "step": 2074 }, { "epoch": 0.64, "learning_rate": 9.172634526905382e-06, "loss": 0.2318, "step": 2075 }, { "epoch": 0.64, "learning_rate": 9.172234446889379e-06, "loss": 0.3515, "step": 2076 }, { "epoch": 0.64, "learning_rate": 9.171834366873375e-06, "loss": 0.4131, "step": 2077 }, { "epoch": 0.64, "learning_rate": 9.171434286857372e-06, "loss": 0.277, "step": 2078 }, { "epoch": 0.65, "learning_rate": 9.171034206841369e-06, "loss": 0.2134, "step": 2079 }, { "epoch": 0.65, "learning_rate": 9.170634126825367e-06, "loss": 0.1102, "step": 2080 }, { "epoch": 0.65, "learning_rate": 9.170234046809364e-06, "loss": 0.2202, "step": 2081 }, { "epoch": 0.65, "learning_rate": 9.169833966793359e-06, "loss": 0.2724, "step": 2082 }, { "epoch": 0.65, "learning_rate": 9.169433886777355e-06, "loss": 0.212, "step": 2083 }, { "epoch": 0.65, "learning_rate": 9.169033806761352e-06, "loss": 0.3178, "step": 2084 }, { "epoch": 0.65, "learning_rate": 9.16863372674535e-06, "loss": 0.3365, "step": 2085 }, { "epoch": 0.65, "learning_rate": 9.168233646729347e-06, "loss": 0.1986, "step": 2086 }, { "epoch": 0.65, "learning_rate": 9.167833566713344e-06, "loss": 0.1775, "step": 2087 }, { "epoch": 0.65, "learning_rate": 9.16743348669734e-06, "loss": 0.1994, "step": 2088 }, { "epoch": 0.65, "learning_rate": 9.167033406681337e-06, "loss": 0.2075, "step": 2089 }, { "epoch": 0.65, "learning_rate": 9.166633326665334e-06, "loss": 0.3002, "step": 2090 }, { "epoch": 0.65, "learning_rate": 9.16623324664933e-06, "loss": 0.1915, "step": 2091 }, { "epoch": 0.65, "learning_rate": 9.165833166633327e-06, "loss": 0.1704, "step": 2092 }, { "epoch": 0.65, "learning_rate": 9.165433086617324e-06, "loss": 0.3366, "step": 2093 }, { "epoch": 0.65, "learning_rate": 9.165033006601322e-06, "loss": 0.3242, "step": 2094 }, { "epoch": 0.65, "learning_rate": 9.164632926585317e-06, "loss": 0.0873, "step": 2095 }, { "epoch": 0.65, "learning_rate": 9.164232846569314e-06, "loss": 0.2541, "step": 2096 }, { "epoch": 0.65, "learning_rate": 9.16383276655331e-06, "loss": 0.3032, "step": 2097 }, { "epoch": 0.65, "learning_rate": 9.163432686537307e-06, "loss": 0.4101, "step": 2098 }, { "epoch": 0.65, "learning_rate": 9.163032606521306e-06, "loss": 0.2296, "step": 2099 }, { "epoch": 0.65, "learning_rate": 9.162632526505302e-06, "loss": 0.2978, "step": 2100 }, { "epoch": 0.65, "learning_rate": 9.162232446489299e-06, "loss": 0.3227, "step": 2101 }, { "epoch": 0.65, "learning_rate": 9.161832366473296e-06, "loss": 0.3309, "step": 2102 }, { "epoch": 0.65, "learning_rate": 9.161432286457292e-06, "loss": 0.2629, "step": 2103 }, { "epoch": 0.65, "learning_rate": 9.161032206441289e-06, "loss": 0.2588, "step": 2104 }, { "epoch": 0.65, "learning_rate": 9.160632126425286e-06, "loss": 0.2271, "step": 2105 }, { "epoch": 0.65, "learning_rate": 9.160232046409282e-06, "loss": 0.1722, "step": 2106 }, { "epoch": 0.65, "learning_rate": 9.15983196639328e-06, "loss": 0.2907, "step": 2107 }, { "epoch": 0.65, "learning_rate": 9.159431886377276e-06, "loss": 0.3194, "step": 2108 }, { "epoch": 0.65, "learning_rate": 9.159031806361273e-06, "loss": 0.2524, "step": 2109 }, { "epoch": 0.65, "learning_rate": 9.15863172634527e-06, "loss": 0.291, "step": 2110 }, { "epoch": 0.66, "learning_rate": 9.158231646329266e-06, "loss": 0.3729, "step": 2111 }, { "epoch": 0.66, "learning_rate": 9.157831566313263e-06, "loss": 0.1725, "step": 2112 }, { "epoch": 0.66, "learning_rate": 9.157431486297261e-06, "loss": 0.2885, "step": 2113 }, { "epoch": 0.66, "learning_rate": 9.157031406281258e-06, "loss": 0.271, "step": 2114 }, { "epoch": 0.66, "learning_rate": 9.156631326265254e-06, "loss": 0.2237, "step": 2115 }, { "epoch": 0.66, "learning_rate": 9.156231246249251e-06, "loss": 0.2874, "step": 2116 }, { "epoch": 0.66, "learning_rate": 9.155831166233248e-06, "loss": 0.3221, "step": 2117 }, { "epoch": 0.66, "learning_rate": 9.155431086217244e-06, "loss": 0.2819, "step": 2118 }, { "epoch": 0.66, "learning_rate": 9.155031006201241e-06, "loss": 0.3109, "step": 2119 }, { "epoch": 0.66, "learning_rate": 9.154630926185238e-06, "loss": 0.284, "step": 2120 }, { "epoch": 0.66, "learning_rate": 9.154230846169234e-06, "loss": 0.2483, "step": 2121 }, { "epoch": 0.66, "learning_rate": 9.153830766153231e-06, "loss": 0.2866, "step": 2122 }, { "epoch": 0.66, "learning_rate": 9.153430686137228e-06, "loss": 0.1534, "step": 2123 }, { "epoch": 0.66, "learning_rate": 9.153030606121225e-06, "loss": 0.2833, "step": 2124 }, { "epoch": 0.66, "learning_rate": 9.152630526105221e-06, "loss": 0.3569, "step": 2125 }, { "epoch": 0.66, "learning_rate": 9.152230446089218e-06, "loss": 0.2594, "step": 2126 }, { "epoch": 0.66, "learning_rate": 9.151830366073216e-06, "loss": 0.2435, "step": 2127 }, { "epoch": 0.66, "learning_rate": 9.151430286057213e-06, "loss": 0.1392, "step": 2128 }, { "epoch": 0.66, "learning_rate": 9.15103020604121e-06, "loss": 0.3822, "step": 2129 }, { "epoch": 0.66, "learning_rate": 9.150630126025206e-06, "loss": 0.3028, "step": 2130 }, { "epoch": 0.66, "learning_rate": 9.150230046009203e-06, "loss": 0.3909, "step": 2131 }, { "epoch": 0.66, "learning_rate": 9.149829965993198e-06, "loss": 0.2895, "step": 2132 }, { "epoch": 0.66, "learning_rate": 9.149429885977196e-06, "loss": 0.3015, "step": 2133 }, { "epoch": 0.66, "learning_rate": 9.149029805961193e-06, "loss": 0.2114, "step": 2134 }, { "epoch": 0.66, "learning_rate": 9.14862972594519e-06, "loss": 0.2117, "step": 2135 }, { "epoch": 0.66, "learning_rate": 9.148229645929186e-06, "loss": 0.3366, "step": 2136 }, { "epoch": 0.66, "learning_rate": 9.147829565913183e-06, "loss": 0.3237, "step": 2137 }, { "epoch": 0.66, "learning_rate": 9.14742948589718e-06, "loss": 0.2196, "step": 2138 }, { "epoch": 0.66, "learning_rate": 9.147029405881177e-06, "loss": 0.3289, "step": 2139 }, { "epoch": 0.66, "learning_rate": 9.146629325865173e-06, "loss": 0.2358, "step": 2140 }, { "epoch": 0.66, "learning_rate": 9.146229245849172e-06, "loss": 0.2802, "step": 2141 }, { "epoch": 0.66, "learning_rate": 9.145829165833168e-06, "loss": 0.2837, "step": 2142 }, { "epoch": 0.67, "learning_rate": 9.145429085817165e-06, "loss": 0.3383, "step": 2143 }, { "epoch": 0.67, "learning_rate": 9.145029005801162e-06, "loss": 0.3026, "step": 2144 }, { "epoch": 0.67, "learning_rate": 9.144628925785157e-06, "loss": 0.1868, "step": 2145 }, { "epoch": 0.67, "learning_rate": 9.144228845769153e-06, "loss": 0.3165, "step": 2146 }, { "epoch": 0.67, "learning_rate": 9.143828765753152e-06, "loss": 0.3381, "step": 2147 }, { "epoch": 0.67, "learning_rate": 9.143428685737148e-06, "loss": 0.2783, "step": 2148 }, { "epoch": 0.67, "learning_rate": 9.143028605721145e-06, "loss": 0.3414, "step": 2149 }, { "epoch": 0.67, "learning_rate": 9.142628525705142e-06, "loss": 0.2963, "step": 2150 }, { "epoch": 0.67, "learning_rate": 9.142228445689138e-06, "loss": 0.2631, "step": 2151 }, { "epoch": 0.67, "learning_rate": 9.141828365673135e-06, "loss": 0.2052, "step": 2152 }, { "epoch": 0.67, "learning_rate": 9.141428285657132e-06, "loss": 0.3342, "step": 2153 }, { "epoch": 0.67, "learning_rate": 9.141028205641129e-06, "loss": 0.2605, "step": 2154 }, { "epoch": 0.67, "learning_rate": 9.140628125625127e-06, "loss": 0.2382, "step": 2155 }, { "epoch": 0.67, "learning_rate": 9.140228045609124e-06, "loss": 0.3674, "step": 2156 }, { "epoch": 0.67, "learning_rate": 9.13982796559312e-06, "loss": 0.4084, "step": 2157 }, { "epoch": 0.67, "learning_rate": 9.139427885577115e-06, "loss": 0.1819, "step": 2158 }, { "epoch": 0.67, "learning_rate": 9.139027805561112e-06, "loss": 0.336, "step": 2159 }, { "epoch": 0.67, "learning_rate": 9.138627725545109e-06, "loss": 0.1945, "step": 2160 }, { "epoch": 0.67, "learning_rate": 9.138227645529107e-06, "loss": 0.3332, "step": 2161 }, { "epoch": 0.67, "learning_rate": 9.137827565513104e-06, "loss": 0.2484, "step": 2162 }, { "epoch": 0.67, "learning_rate": 9.1374274854971e-06, "loss": 0.2275, "step": 2163 }, { "epoch": 0.67, "learning_rate": 9.137027405481097e-06, "loss": 0.2377, "step": 2164 }, { "epoch": 0.67, "learning_rate": 9.136627325465094e-06, "loss": 0.1688, "step": 2165 }, { "epoch": 0.67, "learning_rate": 9.13622724544909e-06, "loss": 0.248, "step": 2166 }, { "epoch": 0.67, "learning_rate": 9.135827165433087e-06, "loss": 0.1959, "step": 2167 }, { "epoch": 0.67, "learning_rate": 9.135427085417084e-06, "loss": 0.2654, "step": 2168 }, { "epoch": 0.67, "learning_rate": 9.135027005401082e-06, "loss": 0.3697, "step": 2169 }, { "epoch": 0.67, "learning_rate": 9.134626925385077e-06, "loss": 0.2285, "step": 2170 }, { "epoch": 0.67, "learning_rate": 9.134226845369074e-06, "loss": 0.3751, "step": 2171 }, { "epoch": 0.67, "learning_rate": 9.13382676535307e-06, "loss": 0.2764, "step": 2172 }, { "epoch": 0.67, "learning_rate": 9.133426685337067e-06, "loss": 0.2558, "step": 2173 }, { "epoch": 0.67, "learning_rate": 9.133026605321066e-06, "loss": 0.4041, "step": 2174 }, { "epoch": 0.68, "learning_rate": 9.132626525305062e-06, "loss": 0.1554, "step": 2175 }, { "epoch": 0.68, "learning_rate": 9.132226445289059e-06, "loss": 0.2457, "step": 2176 }, { "epoch": 0.68, "learning_rate": 9.131826365273056e-06, "loss": 0.318, "step": 2177 }, { "epoch": 0.68, "learning_rate": 9.131426285257052e-06, "loss": 0.3644, "step": 2178 }, { "epoch": 0.68, "learning_rate": 9.131026205241049e-06, "loss": 0.3206, "step": 2179 }, { "epoch": 0.68, "learning_rate": 9.130626125225046e-06, "loss": 0.2958, "step": 2180 }, { "epoch": 0.68, "learning_rate": 9.130226045209042e-06, "loss": 0.1958, "step": 2181 }, { "epoch": 0.68, "learning_rate": 9.129825965193039e-06, "loss": 0.1848, "step": 2182 }, { "epoch": 0.68, "learning_rate": 9.129425885177036e-06, "loss": 0.2527, "step": 2183 }, { "epoch": 0.68, "learning_rate": 9.129025805161032e-06, "loss": 0.1337, "step": 2184 }, { "epoch": 0.68, "learning_rate": 9.12862572514503e-06, "loss": 0.3479, "step": 2185 }, { "epoch": 0.68, "learning_rate": 9.128225645129026e-06, "loss": 0.3363, "step": 2186 }, { "epoch": 0.68, "learning_rate": 9.127825565113023e-06, "loss": 0.3204, "step": 2187 }, { "epoch": 0.68, "learning_rate": 9.127425485097021e-06, "loss": 0.3669, "step": 2188 }, { "epoch": 0.68, "learning_rate": 9.127025405081018e-06, "loss": 0.2258, "step": 2189 }, { "epoch": 0.68, "learning_rate": 9.126625325065014e-06, "loss": 0.2643, "step": 2190 }, { "epoch": 0.68, "learning_rate": 9.126225245049011e-06, "loss": 0.3964, "step": 2191 }, { "epoch": 0.68, "learning_rate": 9.125825165033008e-06, "loss": 0.1672, "step": 2192 }, { "epoch": 0.68, "learning_rate": 9.125425085017004e-06, "loss": 0.2601, "step": 2193 }, { "epoch": 0.68, "learning_rate": 9.125025005001001e-06, "loss": 0.3261, "step": 2194 }, { "epoch": 0.68, "learning_rate": 9.124624924984998e-06, "loss": 0.2845, "step": 2195 }, { "epoch": 0.68, "learning_rate": 9.124224844968994e-06, "loss": 0.2349, "step": 2196 }, { "epoch": 0.68, "learning_rate": 9.123824764952991e-06, "loss": 0.435, "step": 2197 }, { "epoch": 0.68, "learning_rate": 9.123424684936988e-06, "loss": 0.2935, "step": 2198 }, { "epoch": 0.68, "learning_rate": 9.123024604920984e-06, "loss": 0.2924, "step": 2199 }, { "epoch": 0.68, "learning_rate": 9.122624524904981e-06, "loss": 0.2544, "step": 2200 }, { "epoch": 0.68, "learning_rate": 9.122224444888978e-06, "loss": 0.1646, "step": 2201 }, { "epoch": 0.68, "learning_rate": 9.121824364872976e-06, "loss": 0.3272, "step": 2202 }, { "epoch": 0.68, "learning_rate": 9.121424284856973e-06, "loss": 0.2323, "step": 2203 }, { "epoch": 0.68, "learning_rate": 9.12102420484097e-06, "loss": 0.4141, "step": 2204 }, { "epoch": 0.68, "learning_rate": 9.120624124824966e-06, "loss": 0.3648, "step": 2205 }, { "epoch": 0.68, "learning_rate": 9.120224044808963e-06, "loss": 0.2283, "step": 2206 }, { "epoch": 0.69, "learning_rate": 9.11982396479296e-06, "loss": 0.3842, "step": 2207 }, { "epoch": 0.69, "learning_rate": 9.119423884776956e-06, "loss": 0.1252, "step": 2208 }, { "epoch": 0.69, "learning_rate": 9.119023804760953e-06, "loss": 0.2992, "step": 2209 }, { "epoch": 0.69, "learning_rate": 9.11862372474495e-06, "loss": 0.2097, "step": 2210 }, { "epoch": 0.69, "learning_rate": 9.118223644728946e-06, "loss": 0.4676, "step": 2211 }, { "epoch": 0.69, "learning_rate": 9.117823564712943e-06, "loss": 0.3023, "step": 2212 }, { "epoch": 0.69, "learning_rate": 9.11742348469694e-06, "loss": 0.1879, "step": 2213 }, { "epoch": 0.69, "learning_rate": 9.117023404680936e-06, "loss": 0.3778, "step": 2214 }, { "epoch": 0.69, "learning_rate": 9.116623324664933e-06, "loss": 0.1971, "step": 2215 }, { "epoch": 0.69, "learning_rate": 9.116223244648932e-06, "loss": 0.2416, "step": 2216 }, { "epoch": 0.69, "learning_rate": 9.115823164632928e-06, "loss": 0.2763, "step": 2217 }, { "epoch": 0.69, "learning_rate": 9.115423084616925e-06, "loss": 0.388, "step": 2218 }, { "epoch": 0.69, "learning_rate": 9.115023004600922e-06, "loss": 0.2119, "step": 2219 }, { "epoch": 0.69, "learning_rate": 9.114622924584918e-06, "loss": 0.3064, "step": 2220 }, { "epoch": 0.69, "learning_rate": 9.114222844568913e-06, "loss": 0.4348, "step": 2221 }, { "epoch": 0.69, "learning_rate": 9.113822764552912e-06, "loss": 0.3006, "step": 2222 }, { "epoch": 0.69, "learning_rate": 9.113422684536908e-06, "loss": 0.1749, "step": 2223 }, { "epoch": 0.69, "learning_rate": 9.113022604520905e-06, "loss": 0.2783, "step": 2224 }, { "epoch": 0.69, "learning_rate": 9.112622524504902e-06, "loss": 0.3324, "step": 2225 }, { "epoch": 0.69, "learning_rate": 9.112222444488898e-06, "loss": 0.2356, "step": 2226 }, { "epoch": 0.69, "learning_rate": 9.111822364472895e-06, "loss": 0.2568, "step": 2227 }, { "epoch": 0.69, "learning_rate": 9.111422284456892e-06, "loss": 0.2182, "step": 2228 }, { "epoch": 0.69, "learning_rate": 9.111022204440888e-06, "loss": 0.1795, "step": 2229 }, { "epoch": 0.69, "learning_rate": 9.110622124424887e-06, "loss": 0.3231, "step": 2230 }, { "epoch": 0.69, "learning_rate": 9.110222044408884e-06, "loss": 0.2152, "step": 2231 }, { "epoch": 0.69, "learning_rate": 9.10982196439288e-06, "loss": 0.3216, "step": 2232 }, { "epoch": 0.69, "learning_rate": 9.109421884376875e-06, "loss": 0.3895, "step": 2233 }, { "epoch": 0.69, "learning_rate": 9.109021804360872e-06, "loss": 0.1547, "step": 2234 }, { "epoch": 0.69, "learning_rate": 9.108621724344869e-06, "loss": 0.2109, "step": 2235 }, { "epoch": 0.69, "learning_rate": 9.108221644328867e-06, "loss": 0.3348, "step": 2236 }, { "epoch": 0.69, "learning_rate": 9.107821564312864e-06, "loss": 0.3949, "step": 2237 }, { "epoch": 0.69, "learning_rate": 9.10742148429686e-06, "loss": 0.3479, "step": 2238 }, { "epoch": 0.69, "learning_rate": 9.107021404280857e-06, "loss": 0.2825, "step": 2239 }, { "epoch": 0.7, "learning_rate": 9.106621324264854e-06, "loss": 0.2312, "step": 2240 }, { "epoch": 0.7, "learning_rate": 9.10622124424885e-06, "loss": 0.2451, "step": 2241 }, { "epoch": 0.7, "learning_rate": 9.105821164232847e-06, "loss": 0.1586, "step": 2242 }, { "epoch": 0.7, "learning_rate": 9.105421084216844e-06, "loss": 0.3648, "step": 2243 }, { "epoch": 0.7, "learning_rate": 9.105021004200842e-06, "loss": 0.1887, "step": 2244 }, { "epoch": 0.7, "learning_rate": 9.104620924184839e-06, "loss": 0.279, "step": 2245 }, { "epoch": 0.7, "learning_rate": 9.104220844168834e-06, "loss": 0.3109, "step": 2246 }, { "epoch": 0.7, "learning_rate": 9.10382076415283e-06, "loss": 0.1887, "step": 2247 }, { "epoch": 0.7, "learning_rate": 9.103420684136827e-06, "loss": 0.1613, "step": 2248 }, { "epoch": 0.7, "learning_rate": 9.103020604120824e-06, "loss": 0.3459, "step": 2249 }, { "epoch": 0.7, "learning_rate": 9.102620524104822e-06, "loss": 0.3185, "step": 2250 }, { "epoch": 0.7, "learning_rate": 9.102220444088819e-06, "loss": 0.3512, "step": 2251 }, { "epoch": 0.7, "learning_rate": 9.101820364072816e-06, "loss": 0.1772, "step": 2252 }, { "epoch": 0.7, "learning_rate": 9.101420284056812e-06, "loss": 0.1939, "step": 2253 }, { "epoch": 0.7, "learning_rate": 9.101020204040809e-06, "loss": 0.2109, "step": 2254 }, { "epoch": 0.7, "learning_rate": 9.100620124024806e-06, "loss": 0.2344, "step": 2255 }, { "epoch": 0.7, "learning_rate": 9.100220044008802e-06, "loss": 0.205, "step": 2256 }, { "epoch": 0.7, "learning_rate": 9.099819963992799e-06, "loss": 0.2792, "step": 2257 }, { "epoch": 0.7, "learning_rate": 9.099419883976797e-06, "loss": 0.2678, "step": 2258 }, { "epoch": 0.7, "learning_rate": 9.099019803960792e-06, "loss": 0.1381, "step": 2259 }, { "epoch": 0.7, "learning_rate": 9.098619723944789e-06, "loss": 0.3859, "step": 2260 }, { "epoch": 0.7, "learning_rate": 9.098219643928786e-06, "loss": 0.2878, "step": 2261 }, { "epoch": 0.7, "learning_rate": 9.097819563912782e-06, "loss": 0.1312, "step": 2262 }, { "epoch": 0.7, "learning_rate": 9.09741948389678e-06, "loss": 0.2598, "step": 2263 }, { "epoch": 0.7, "learning_rate": 9.097019403880778e-06, "loss": 0.2349, "step": 2264 }, { "epoch": 0.7, "learning_rate": 9.096619323864774e-06, "loss": 0.3033, "step": 2265 }, { "epoch": 0.7, "learning_rate": 9.096219243848771e-06, "loss": 0.2247, "step": 2266 }, { "epoch": 0.7, "learning_rate": 9.095819163832768e-06, "loss": 0.2439, "step": 2267 }, { "epoch": 0.7, "learning_rate": 9.095419083816764e-06, "loss": 0.2534, "step": 2268 }, { "epoch": 0.7, "learning_rate": 9.095019003800761e-06, "loss": 0.2662, "step": 2269 }, { "epoch": 0.7, "learning_rate": 9.094618923784758e-06, "loss": 0.2621, "step": 2270 }, { "epoch": 0.7, "learning_rate": 9.094218843768754e-06, "loss": 0.2698, "step": 2271 }, { "epoch": 0.71, "learning_rate": 9.093818763752751e-06, "loss": 0.2689, "step": 2272 }, { "epoch": 0.71, "learning_rate": 9.093418683736748e-06, "loss": 0.4508, "step": 2273 }, { "epoch": 0.71, "learning_rate": 9.093018603720744e-06, "loss": 0.2998, "step": 2274 }, { "epoch": 0.71, "learning_rate": 9.092618523704741e-06, "loss": 0.2591, "step": 2275 }, { "epoch": 0.71, "learning_rate": 9.092218443688738e-06, "loss": 0.2859, "step": 2276 }, { "epoch": 0.71, "learning_rate": 9.091818363672734e-06, "loss": 0.2355, "step": 2277 }, { "epoch": 0.71, "learning_rate": 9.091418283656733e-06, "loss": 0.1599, "step": 2278 }, { "epoch": 0.71, "learning_rate": 9.09101820364073e-06, "loss": 0.2, "step": 2279 }, { "epoch": 0.71, "learning_rate": 9.090618123624726e-06, "loss": 0.2861, "step": 2280 }, { "epoch": 0.71, "learning_rate": 9.090218043608723e-06, "loss": 0.168, "step": 2281 }, { "epoch": 0.71, "learning_rate": 9.08981796359272e-06, "loss": 0.1756, "step": 2282 }, { "epoch": 0.71, "learning_rate": 9.089417883576716e-06, "loss": 0.6802, "step": 2283 }, { "epoch": 0.71, "learning_rate": 9.089017803560713e-06, "loss": 0.492, "step": 2284 }, { "epoch": 0.71, "learning_rate": 9.08861772354471e-06, "loss": 0.2945, "step": 2285 }, { "epoch": 0.71, "learning_rate": 9.088217643528706e-06, "loss": 0.1961, "step": 2286 }, { "epoch": 0.71, "learning_rate": 9.087817563512703e-06, "loss": 0.196, "step": 2287 }, { "epoch": 0.71, "learning_rate": 9.0874174834967e-06, "loss": 0.1698, "step": 2288 }, { "epoch": 0.71, "learning_rate": 9.0874174834967e-06, "loss": 0.4307, "step": 2289 }, { "epoch": 0.71, "learning_rate": 9.087017403480696e-06, "loss": 0.174, "step": 2290 }, { "epoch": 0.71, "learning_rate": 9.086617323464693e-06, "loss": 0.238, "step": 2291 }, { "epoch": 0.71, "learning_rate": 9.08621724344869e-06, "loss": 0.2007, "step": 2292 }, { "epoch": 0.71, "learning_rate": 9.085817163432688e-06, "loss": 0.3148, "step": 2293 }, { "epoch": 0.71, "learning_rate": 9.085417083416685e-06, "loss": 0.3631, "step": 2294 }, { "epoch": 0.71, "learning_rate": 9.085017003400682e-06, "loss": 0.2728, "step": 2295 }, { "epoch": 0.71, "learning_rate": 9.084616923384678e-06, "loss": 0.2059, "step": 2296 }, { "epoch": 0.71, "learning_rate": 9.084216843368673e-06, "loss": 0.2204, "step": 2297 }, { "epoch": 0.71, "learning_rate": 9.083816763352672e-06, "loss": 0.2433, "step": 2298 }, { "epoch": 0.71, "learning_rate": 9.083416683336668e-06, "loss": 0.2703, "step": 2299 }, { "epoch": 0.71, "learning_rate": 9.083016603320665e-06, "loss": 0.2091, "step": 2300 }, { "epoch": 0.71, "learning_rate": 9.082616523304662e-06, "loss": 0.2338, "step": 2301 }, { "epoch": 0.71, "learning_rate": 9.082216443288658e-06, "loss": 0.3777, "step": 2302 }, { "epoch": 0.71, "learning_rate": 9.081816363272655e-06, "loss": 0.3453, "step": 2303 }, { "epoch": 0.72, "learning_rate": 9.081416283256652e-06, "loss": 0.2725, "step": 2304 }, { "epoch": 0.72, "learning_rate": 9.081016203240648e-06, "loss": 0.2874, "step": 2305 }, { "epoch": 0.72, "learning_rate": 9.080616123224645e-06, "loss": 0.2429, "step": 2306 }, { "epoch": 0.72, "learning_rate": 9.080216043208643e-06, "loss": 0.2159, "step": 2307 }, { "epoch": 0.72, "learning_rate": 9.07981596319264e-06, "loss": 0.3018, "step": 2308 }, { "epoch": 0.72, "learning_rate": 9.079415883176637e-06, "loss": 0.1979, "step": 2309 }, { "epoch": 0.72, "learning_rate": 9.079015803160632e-06, "loss": 0.4501, "step": 2310 }, { "epoch": 0.72, "learning_rate": 9.078615723144629e-06, "loss": 0.1503, "step": 2311 }, { "epoch": 0.72, "learning_rate": 9.078215643128627e-06, "loss": 0.2667, "step": 2312 }, { "epoch": 0.72, "learning_rate": 9.077815563112624e-06, "loss": 0.2322, "step": 2313 }, { "epoch": 0.72, "learning_rate": 9.07741548309662e-06, "loss": 0.1382, "step": 2314 }, { "epoch": 0.72, "learning_rate": 9.077015403080617e-06, "loss": 0.3367, "step": 2315 }, { "epoch": 0.72, "learning_rate": 9.076615323064614e-06, "loss": 0.2379, "step": 2316 }, { "epoch": 0.72, "learning_rate": 9.07621524304861e-06, "loss": 0.1787, "step": 2317 }, { "epoch": 0.72, "learning_rate": 9.075815163032607e-06, "loss": 0.0817, "step": 2318 }, { "epoch": 0.72, "learning_rate": 9.075415083016604e-06, "loss": 0.2584, "step": 2319 }, { "epoch": 0.72, "learning_rate": 9.075015003000602e-06, "loss": 0.1597, "step": 2320 }, { "epoch": 0.72, "learning_rate": 9.074614922984599e-06, "loss": 0.1965, "step": 2321 }, { "epoch": 0.72, "learning_rate": 9.074214842968595e-06, "loss": 0.4168, "step": 2322 }, { "epoch": 0.72, "learning_rate": 9.07381476295259e-06, "loss": 0.2707, "step": 2323 }, { "epoch": 0.72, "learning_rate": 9.073414682936587e-06, "loss": 0.3916, "step": 2324 }, { "epoch": 0.72, "learning_rate": 9.073014602920584e-06, "loss": 0.247, "step": 2325 }, { "epoch": 0.72, "learning_rate": 9.072614522904582e-06, "loss": 0.2563, "step": 2326 }, { "epoch": 0.72, "learning_rate": 9.072214442888579e-06, "loss": 0.3853, "step": 2327 }, { "epoch": 0.72, "learning_rate": 9.071814362872576e-06, "loss": 0.2201, "step": 2328 }, { "epoch": 0.72, "learning_rate": 9.071414282856572e-06, "loss": 0.1829, "step": 2329 }, { "epoch": 0.72, "learning_rate": 9.071014202840569e-06, "loss": 0.2782, "step": 2330 }, { "epoch": 0.72, "learning_rate": 9.070614122824566e-06, "loss": 0.2784, "step": 2331 }, { "epoch": 0.72, "learning_rate": 9.070214042808562e-06, "loss": 0.2043, "step": 2332 }, { "epoch": 0.72, "learning_rate": 9.069813962792559e-06, "loss": 0.2588, "step": 2333 }, { "epoch": 0.72, "learning_rate": 9.069413882776557e-06, "loss": 0.3856, "step": 2334 }, { "epoch": 0.72, "learning_rate": 9.069013802760552e-06, "loss": 0.239, "step": 2335 }, { "epoch": 0.73, "learning_rate": 9.068613722744549e-06, "loss": 0.2035, "step": 2336 }, { "epoch": 0.73, "learning_rate": 9.068213642728546e-06, "loss": 0.3207, "step": 2337 }, { "epoch": 0.73, "learning_rate": 9.067813562712542e-06, "loss": 0.2694, "step": 2338 }, { "epoch": 0.73, "learning_rate": 9.067413482696539e-06, "loss": 0.2259, "step": 2339 }, { "epoch": 0.73, "learning_rate": 9.067013402680537e-06, "loss": 0.2252, "step": 2340 }, { "epoch": 0.73, "learning_rate": 9.066613322664534e-06, "loss": 0.4582, "step": 2341 }, { "epoch": 0.73, "learning_rate": 9.066213242648531e-06, "loss": 0.3262, "step": 2342 }, { "epoch": 0.73, "learning_rate": 9.065813162632528e-06, "loss": 0.2492, "step": 2343 }, { "epoch": 0.73, "learning_rate": 9.065413082616524e-06, "loss": 0.2929, "step": 2344 }, { "epoch": 0.73, "learning_rate": 9.065013002600521e-06, "loss": 0.2664, "step": 2345 }, { "epoch": 0.73, "learning_rate": 9.064612922584518e-06, "loss": 0.242, "step": 2346 }, { "epoch": 0.73, "learning_rate": 9.064212842568514e-06, "loss": 0.2413, "step": 2347 }, { "epoch": 0.73, "learning_rate": 9.063812762552511e-06, "loss": 0.209, "step": 2348 }, { "epoch": 0.73, "learning_rate": 9.063412682536508e-06, "loss": 0.2545, "step": 2349 }, { "epoch": 0.73, "learning_rate": 9.063012602520504e-06, "loss": 0.2005, "step": 2350 }, { "epoch": 0.73, "learning_rate": 9.062612522504501e-06, "loss": 0.2802, "step": 2351 }, { "epoch": 0.73, "learning_rate": 9.062212442488498e-06, "loss": 0.3753, "step": 2352 }, { "epoch": 0.73, "learning_rate": 9.061812362472494e-06, "loss": 0.2992, "step": 2353 }, { "epoch": 0.73, "learning_rate": 9.061412282456493e-06, "loss": 0.2592, "step": 2354 }, { "epoch": 0.73, "learning_rate": 9.06101220244049e-06, "loss": 0.2959, "step": 2355 }, { "epoch": 0.73, "learning_rate": 9.060612122424486e-06, "loss": 0.2077, "step": 2356 }, { "epoch": 0.73, "learning_rate": 9.060212042408483e-06, "loss": 0.1913, "step": 2357 }, { "epoch": 0.73, "learning_rate": 9.05981196239248e-06, "loss": 0.4812, "step": 2358 }, { "epoch": 0.73, "learning_rate": 9.059411882376476e-06, "loss": 0.2444, "step": 2359 }, { "epoch": 0.73, "learning_rate": 9.059011802360473e-06, "loss": 0.2366, "step": 2360 }, { "epoch": 0.73, "learning_rate": 9.05861172234447e-06, "loss": 0.3798, "step": 2361 }, { "epoch": 0.73, "learning_rate": 9.058211642328466e-06, "loss": 0.1671, "step": 2362 }, { "epoch": 0.73, "learning_rate": 9.057811562312463e-06, "loss": 0.2084, "step": 2363 }, { "epoch": 0.73, "learning_rate": 9.05741148229646e-06, "loss": 0.2769, "step": 2364 }, { "epoch": 0.73, "learning_rate": 9.057011402280456e-06, "loss": 0.305, "step": 2365 }, { "epoch": 0.73, "learning_rate": 9.056611322264453e-06, "loss": 0.2161, "step": 2366 }, { "epoch": 0.73, "learning_rate": 9.05621124224845e-06, "loss": 0.3372, "step": 2367 }, { "epoch": 0.73, "learning_rate": 9.055811162232448e-06, "loss": 0.2528, "step": 2368 }, { "epoch": 0.74, "learning_rate": 9.055411082216445e-06, "loss": 0.1769, "step": 2369 }, { "epoch": 0.74, "learning_rate": 9.055011002200441e-06, "loss": 0.1792, "step": 2370 }, { "epoch": 0.74, "learning_rate": 9.054610922184438e-06, "loss": 0.238, "step": 2371 }, { "epoch": 0.74, "learning_rate": 9.054210842168435e-06, "loss": 0.5193, "step": 2372 }, { "epoch": 0.74, "learning_rate": 9.053810762152432e-06, "loss": 0.2091, "step": 2373 }, { "epoch": 0.74, "learning_rate": 9.053410682136428e-06, "loss": 0.22, "step": 2374 }, { "epoch": 0.74, "learning_rate": 9.053010602120425e-06, "loss": 0.4759, "step": 2375 }, { "epoch": 0.74, "learning_rate": 9.052610522104422e-06, "loss": 0.2864, "step": 2376 }, { "epoch": 0.74, "learning_rate": 9.052210442088418e-06, "loss": 0.2857, "step": 2377 }, { "epoch": 0.74, "learning_rate": 9.051810362072415e-06, "loss": 0.1797, "step": 2378 }, { "epoch": 0.74, "learning_rate": 9.051410282056412e-06, "loss": 0.281, "step": 2379 }, { "epoch": 0.74, "learning_rate": 9.051010202040408e-06, "loss": 0.2109, "step": 2380 }, { "epoch": 0.74, "learning_rate": 9.050610122024405e-06, "loss": 0.4907, "step": 2381 }, { "epoch": 0.74, "learning_rate": 9.050210042008403e-06, "loss": 0.4006, "step": 2382 }, { "epoch": 0.74, "learning_rate": 9.0498099619924e-06, "loss": 0.319, "step": 2383 }, { "epoch": 0.74, "learning_rate": 9.049409881976397e-06, "loss": 0.2521, "step": 2384 }, { "epoch": 0.74, "learning_rate": 9.049009801960393e-06, "loss": 0.3147, "step": 2385 }, { "epoch": 0.74, "learning_rate": 9.048609721944388e-06, "loss": 0.2265, "step": 2386 }, { "epoch": 0.74, "learning_rate": 9.048209641928387e-06, "loss": 0.226, "step": 2387 }, { "epoch": 0.74, "learning_rate": 9.047809561912384e-06, "loss": 0.276, "step": 2388 }, { "epoch": 0.74, "learning_rate": 9.04740948189638e-06, "loss": 0.2718, "step": 2389 }, { "epoch": 0.74, "learning_rate": 9.047009401880377e-06, "loss": 0.3103, "step": 2390 }, { "epoch": 0.74, "learning_rate": 9.046609321864374e-06, "loss": 0.23, "step": 2391 }, { "epoch": 0.74, "learning_rate": 9.04620924184837e-06, "loss": 0.2358, "step": 2392 }, { "epoch": 0.74, "learning_rate": 9.045809161832367e-06, "loss": 0.308, "step": 2393 }, { "epoch": 0.74, "learning_rate": 9.045409081816364e-06, "loss": 0.2066, "step": 2394 }, { "epoch": 0.74, "learning_rate": 9.04500900180036e-06, "loss": 0.3788, "step": 2395 }, { "epoch": 0.74, "learning_rate": 9.044608921784359e-06, "loss": 0.3186, "step": 2396 }, { "epoch": 0.74, "learning_rate": 9.044208841768355e-06, "loss": 0.2742, "step": 2397 }, { "epoch": 0.74, "learning_rate": 9.04380876175235e-06, "loss": 0.2699, "step": 2398 }, { "epoch": 0.74, "learning_rate": 9.043408681736347e-06, "loss": 0.2611, "step": 2399 }, { "epoch": 0.74, "learning_rate": 9.043008601720344e-06, "loss": 0.4024, "step": 2400 }, { "epoch": 0.75, "learning_rate": 9.042608521704342e-06, "loss": 0.2664, "step": 2401 }, { "epoch": 0.75, "learning_rate": 9.042208441688339e-06, "loss": 0.242, "step": 2402 }, { "epoch": 0.75, "learning_rate": 9.041808361672336e-06, "loss": 0.2372, "step": 2403 }, { "epoch": 0.75, "learning_rate": 9.041408281656332e-06, "loss": 0.2512, "step": 2404 }, { "epoch": 0.75, "learning_rate": 9.041008201640329e-06, "loss": 0.3864, "step": 2405 }, { "epoch": 0.75, "learning_rate": 9.040608121624326e-06, "loss": 0.2633, "step": 2406 }, { "epoch": 0.75, "learning_rate": 9.040208041608322e-06, "loss": 0.1831, "step": 2407 }, { "epoch": 0.75, "learning_rate": 9.039807961592319e-06, "loss": 0.2095, "step": 2408 }, { "epoch": 0.75, "learning_rate": 9.039407881576316e-06, "loss": 0.1918, "step": 2409 }, { "epoch": 0.75, "learning_rate": 9.039007801560314e-06, "loss": 0.259, "step": 2410 }, { "epoch": 0.75, "learning_rate": 9.038607721544309e-06, "loss": 0.5429, "step": 2411 }, { "epoch": 0.75, "learning_rate": 9.038207641528306e-06, "loss": 0.2076, "step": 2412 }, { "epoch": 0.75, "learning_rate": 9.037807561512302e-06, "loss": 0.3085, "step": 2413 }, { "epoch": 0.75, "learning_rate": 9.037407481496299e-06, "loss": 0.2887, "step": 2414 }, { "epoch": 0.75, "learning_rate": 9.037007401480297e-06, "loss": 0.2265, "step": 2415 }, { "epoch": 0.75, "learning_rate": 9.036607321464294e-06, "loss": 0.2648, "step": 2416 }, { "epoch": 0.75, "learning_rate": 9.03620724144829e-06, "loss": 0.2155, "step": 2417 }, { "epoch": 0.75, "learning_rate": 9.035807161432287e-06, "loss": 0.1872, "step": 2418 }, { "epoch": 0.75, "learning_rate": 9.035407081416284e-06, "loss": 0.2044, "step": 2419 }, { "epoch": 0.75, "learning_rate": 9.035007001400281e-06, "loss": 0.201, "step": 2420 }, { "epoch": 0.75, "learning_rate": 9.034606921384278e-06, "loss": 0.2333, "step": 2421 }, { "epoch": 0.75, "learning_rate": 9.034206841368274e-06, "loss": 0.2422, "step": 2422 }, { "epoch": 0.75, "learning_rate": 9.033806761352271e-06, "loss": 0.6415, "step": 2423 }, { "epoch": 0.75, "learning_rate": 9.033406681336268e-06, "loss": 0.1477, "step": 2424 }, { "epoch": 0.75, "learning_rate": 9.033006601320264e-06, "loss": 0.1631, "step": 2425 }, { "epoch": 0.75, "learning_rate": 9.032606521304261e-06, "loss": 0.2176, "step": 2426 }, { "epoch": 0.75, "learning_rate": 9.032206441288258e-06, "loss": 0.1456, "step": 2427 }, { "epoch": 0.75, "learning_rate": 9.031806361272254e-06, "loss": 0.1471, "step": 2428 }, { "epoch": 0.75, "learning_rate": 9.031406281256253e-06, "loss": 0.2422, "step": 2429 }, { "epoch": 0.75, "learning_rate": 9.03100620124025e-06, "loss": 0.247, "step": 2430 }, { "epoch": 0.75, "learning_rate": 9.030606121224246e-06, "loss": 0.2773, "step": 2431 }, { "epoch": 0.75, "learning_rate": 9.030206041208243e-06, "loss": 0.1865, "step": 2432 }, { "epoch": 0.76, "learning_rate": 9.02980596119224e-06, "loss": 0.2848, "step": 2433 }, { "epoch": 0.76, "learning_rate": 9.029405881176236e-06, "loss": 0.2354, "step": 2434 }, { "epoch": 0.76, "learning_rate": 9.029005801160233e-06, "loss": 0.1585, "step": 2435 }, { "epoch": 0.76, "learning_rate": 9.02860572114423e-06, "loss": 0.1797, "step": 2436 }, { "epoch": 0.76, "learning_rate": 9.028205641128226e-06, "loss": 0.2366, "step": 2437 }, { "epoch": 0.76, "learning_rate": 9.027805561112223e-06, "loss": 0.2498, "step": 2438 }, { "epoch": 0.76, "learning_rate": 9.02740548109622e-06, "loss": 0.1033, "step": 2439 }, { "epoch": 0.76, "learning_rate": 9.027005401080216e-06, "loss": 0.3468, "step": 2440 }, { "epoch": 0.76, "learning_rate": 9.026605321064213e-06, "loss": 0.3159, "step": 2441 }, { "epoch": 0.76, "learning_rate": 9.02620524104821e-06, "loss": 0.2065, "step": 2442 }, { "epoch": 0.76, "learning_rate": 9.025805161032208e-06, "loss": 0.2867, "step": 2443 }, { "epoch": 0.76, "learning_rate": 9.025405081016205e-06, "loss": 0.2226, "step": 2444 }, { "epoch": 0.76, "learning_rate": 9.025005001000201e-06, "loss": 0.2403, "step": 2445 }, { "epoch": 0.76, "learning_rate": 9.024604920984198e-06, "loss": 0.1724, "step": 2446 }, { "epoch": 0.76, "learning_rate": 9.024204840968195e-06, "loss": 0.1417, "step": 2447 }, { "epoch": 0.76, "learning_rate": 9.023804760952191e-06, "loss": 0.2213, "step": 2448 }, { "epoch": 0.76, "learning_rate": 9.023404680936188e-06, "loss": 0.2532, "step": 2449 }, { "epoch": 0.76, "learning_rate": 9.023004600920185e-06, "loss": 0.2394, "step": 2450 }, { "epoch": 0.76, "learning_rate": 9.022604520904182e-06, "loss": 0.2822, "step": 2451 }, { "epoch": 0.76, "learning_rate": 9.022204440888178e-06, "loss": 0.274, "step": 2452 }, { "epoch": 0.76, "learning_rate": 9.021804360872175e-06, "loss": 0.3318, "step": 2453 }, { "epoch": 0.76, "learning_rate": 9.021404280856172e-06, "loss": 0.1559, "step": 2454 }, { "epoch": 0.76, "learning_rate": 9.021004200840168e-06, "loss": 0.3192, "step": 2455 }, { "epoch": 0.76, "learning_rate": 9.020604120824165e-06, "loss": 0.2672, "step": 2456 }, { "epoch": 0.76, "learning_rate": 9.020204040808163e-06, "loss": 0.2235, "step": 2457 }, { "epoch": 0.76, "learning_rate": 9.01980396079216e-06, "loss": 0.2251, "step": 2458 }, { "epoch": 0.76, "learning_rate": 9.019403880776157e-06, "loss": 0.32, "step": 2459 }, { "epoch": 0.76, "learning_rate": 9.019003800760153e-06, "loss": 0.2646, "step": 2460 }, { "epoch": 0.76, "learning_rate": 9.018603720744148e-06, "loss": 0.2042, "step": 2461 }, { "epoch": 0.76, "learning_rate": 9.018203640728145e-06, "loss": 0.2478, "step": 2462 }, { "epoch": 0.76, "learning_rate": 9.017803560712143e-06, "loss": 0.1517, "step": 2463 }, { "epoch": 0.76, "learning_rate": 9.01740348069614e-06, "loss": 0.3454, "step": 2464 }, { "epoch": 0.77, "learning_rate": 9.017003400680137e-06, "loss": 0.4359, "step": 2465 }, { "epoch": 0.77, "learning_rate": 9.016603320664134e-06, "loss": 0.2723, "step": 2466 }, { "epoch": 0.77, "learning_rate": 9.01620324064813e-06, "loss": 0.1966, "step": 2467 }, { "epoch": 0.77, "learning_rate": 9.015803160632127e-06, "loss": 0.3191, "step": 2468 }, { "epoch": 0.77, "learning_rate": 9.015403080616124e-06, "loss": 0.3087, "step": 2469 }, { "epoch": 0.77, "learning_rate": 9.01500300060012e-06, "loss": 0.2885, "step": 2470 }, { "epoch": 0.77, "learning_rate": 9.014602920584119e-06, "loss": 0.4591, "step": 2471 }, { "epoch": 0.77, "learning_rate": 9.014202840568115e-06, "loss": 0.2182, "step": 2472 }, { "epoch": 0.77, "learning_rate": 9.013802760552112e-06, "loss": 0.2096, "step": 2473 }, { "epoch": 0.77, "learning_rate": 9.013402680536107e-06, "loss": 0.2688, "step": 2474 }, { "epoch": 0.77, "learning_rate": 9.013002600520104e-06, "loss": 0.1956, "step": 2475 }, { "epoch": 0.77, "learning_rate": 9.012602520504102e-06, "loss": 0.2503, "step": 2476 }, { "epoch": 0.77, "learning_rate": 9.012202440488099e-06, "loss": 0.2488, "step": 2477 }, { "epoch": 0.77, "learning_rate": 9.011802360472095e-06, "loss": 0.2531, "step": 2478 }, { "epoch": 0.77, "learning_rate": 9.011402280456092e-06, "loss": 0.3256, "step": 2479 }, { "epoch": 0.77, "learning_rate": 9.011002200440089e-06, "loss": 0.1587, "step": 2480 }, { "epoch": 0.77, "learning_rate": 9.010602120424086e-06, "loss": 0.2706, "step": 2481 }, { "epoch": 0.77, "learning_rate": 9.010202040408082e-06, "loss": 0.3174, "step": 2482 }, { "epoch": 0.77, "learning_rate": 9.009801960392079e-06, "loss": 0.2345, "step": 2483 }, { "epoch": 0.77, "learning_rate": 9.009401880376076e-06, "loss": 0.1687, "step": 2484 }, { "epoch": 0.77, "learning_rate": 9.009001800360074e-06, "loss": 0.1871, "step": 2485 }, { "epoch": 0.77, "learning_rate": 9.00860172034407e-06, "loss": 0.328, "step": 2486 }, { "epoch": 0.77, "learning_rate": 9.008201640328066e-06, "loss": 0.2222, "step": 2487 }, { "epoch": 0.77, "learning_rate": 9.007801560312062e-06, "loss": 0.2275, "step": 2488 }, { "epoch": 0.77, "learning_rate": 9.007401480296059e-06, "loss": 0.4103, "step": 2489 }, { "epoch": 0.77, "learning_rate": 9.007001400280057e-06, "loss": 0.1742, "step": 2490 }, { "epoch": 0.77, "learning_rate": 9.006601320264054e-06, "loss": 0.3993, "step": 2491 }, { "epoch": 0.77, "learning_rate": 9.00620124024805e-06, "loss": 0.2235, "step": 2492 }, { "epoch": 0.77, "learning_rate": 9.005801160232047e-06, "loss": 0.2865, "step": 2493 }, { "epoch": 0.77, "learning_rate": 9.005401080216044e-06, "loss": 0.221, "step": 2494 }, { "epoch": 0.77, "learning_rate": 9.00500100020004e-06, "loss": 0.1951, "step": 2495 }, { "epoch": 0.77, "learning_rate": 9.004600920184037e-06, "loss": 0.3533, "step": 2496 }, { "epoch": 0.78, "learning_rate": 9.004200840168034e-06, "loss": 0.2419, "step": 2497 }, { "epoch": 0.78, "learning_rate": 9.003800760152031e-06, "loss": 0.2413, "step": 2498 }, { "epoch": 0.78, "learning_rate": 9.003400680136028e-06, "loss": 0.1196, "step": 2499 }, { "epoch": 0.78, "learning_rate": 9.003000600120024e-06, "loss": 0.1732, "step": 2500 }, { "epoch": 0.78, "learning_rate": 9.002600520104021e-06, "loss": 0.3128, "step": 2501 }, { "epoch": 0.78, "learning_rate": 9.002200440088018e-06, "loss": 0.2563, "step": 2502 }, { "epoch": 0.78, "learning_rate": 9.001800360072014e-06, "loss": 0.1883, "step": 2503 }, { "epoch": 0.78, "learning_rate": 9.001400280056013e-06, "loss": 0.33, "step": 2504 }, { "epoch": 0.78, "learning_rate": 9.00100020004001e-06, "loss": 0.2152, "step": 2505 }, { "epoch": 0.78, "learning_rate": 9.000600120024006e-06, "loss": 0.2173, "step": 2506 }, { "epoch": 0.78, "learning_rate": 9.000200040008003e-06, "loss": 0.2346, "step": 2507 }, { "epoch": 0.78, "learning_rate": 8.999799959992e-06, "loss": 0.2635, "step": 2508 }, { "epoch": 0.78, "learning_rate": 8.999399879975996e-06, "loss": 0.2512, "step": 2509 }, { "epoch": 0.78, "learning_rate": 8.998999799959993e-06, "loss": 0.1346, "step": 2510 }, { "epoch": 0.78, "learning_rate": 8.99859971994399e-06, "loss": 0.443, "step": 2511 }, { "epoch": 0.78, "learning_rate": 8.998199639927986e-06, "loss": 0.1745, "step": 2512 }, { "epoch": 0.78, "learning_rate": 8.997799559911983e-06, "loss": 0.2988, "step": 2513 }, { "epoch": 0.78, "learning_rate": 8.99739947989598e-06, "loss": 0.372, "step": 2514 }, { "epoch": 0.78, "learning_rate": 8.996999399879976e-06, "loss": 0.2304, "step": 2515 }, { "epoch": 0.78, "learning_rate": 8.996599319863973e-06, "loss": 0.2298, "step": 2516 }, { "epoch": 0.78, "learning_rate": 8.99619923984797e-06, "loss": 0.2671, "step": 2517 }, { "epoch": 0.78, "learning_rate": 8.995799159831968e-06, "loss": 0.2571, "step": 2518 }, { "epoch": 0.78, "learning_rate": 8.995399079815965e-06, "loss": 0.3173, "step": 2519 }, { "epoch": 0.78, "learning_rate": 8.994998999799961e-06, "loss": 0.3839, "step": 2520 }, { "epoch": 0.78, "learning_rate": 8.994598919783958e-06, "loss": 0.2727, "step": 2521 }, { "epoch": 0.78, "learning_rate": 8.994198839767955e-06, "loss": 0.3135, "step": 2522 }, { "epoch": 0.78, "learning_rate": 8.993798759751951e-06, "loss": 0.3739, "step": 2523 }, { "epoch": 0.78, "learning_rate": 8.993398679735948e-06, "loss": 0.2903, "step": 2524 }, { "epoch": 0.78, "learning_rate": 8.992998599719945e-06, "loss": 0.2326, "step": 2525 }, { "epoch": 0.78, "learning_rate": 8.992598519703941e-06, "loss": 0.2227, "step": 2526 }, { "epoch": 0.78, "learning_rate": 8.992198439687938e-06, "loss": 0.2915, "step": 2527 }, { "epoch": 0.78, "learning_rate": 8.991798359671935e-06, "loss": 0.1753, "step": 2528 }, { "epoch": 0.78, "learning_rate": 8.991398279655932e-06, "loss": 0.2606, "step": 2529 }, { "epoch": 0.79, "learning_rate": 8.990998199639928e-06, "loss": 0.2039, "step": 2530 }, { "epoch": 0.79, "learning_rate": 8.990598119623925e-06, "loss": 0.2957, "step": 2531 }, { "epoch": 0.79, "learning_rate": 8.990198039607923e-06, "loss": 0.269, "step": 2532 }, { "epoch": 0.79, "learning_rate": 8.98979795959192e-06, "loss": 0.15, "step": 2533 }, { "epoch": 0.79, "learning_rate": 8.989397879575917e-06, "loss": 0.1618, "step": 2534 }, { "epoch": 0.79, "learning_rate": 8.988997799559913e-06, "loss": 0.1627, "step": 2535 }, { "epoch": 0.79, "learning_rate": 8.98859771954391e-06, "loss": 0.2034, "step": 2536 }, { "epoch": 0.79, "learning_rate": 8.988197639527905e-06, "loss": 0.2097, "step": 2537 }, { "epoch": 0.79, "learning_rate": 8.987797559511903e-06, "loss": 0.3473, "step": 2538 }, { "epoch": 0.79, "learning_rate": 8.9873974794959e-06, "loss": 0.2838, "step": 2539 }, { "epoch": 0.79, "learning_rate": 8.986997399479897e-06, "loss": 0.3448, "step": 2540 }, { "epoch": 0.79, "learning_rate": 8.986597319463893e-06, "loss": 0.3191, "step": 2541 }, { "epoch": 0.79, "learning_rate": 8.98619723944789e-06, "loss": 0.2188, "step": 2542 }, { "epoch": 0.79, "learning_rate": 8.985797159431887e-06, "loss": 0.2197, "step": 2543 }, { "epoch": 0.79, "learning_rate": 8.985397079415884e-06, "loss": 0.3618, "step": 2544 }, { "epoch": 0.79, "learning_rate": 8.98499699939988e-06, "loss": 0.2199, "step": 2545 }, { "epoch": 0.79, "learning_rate": 8.984596919383879e-06, "loss": 0.2629, "step": 2546 }, { "epoch": 0.79, "learning_rate": 8.984196839367875e-06, "loss": 0.2811, "step": 2547 }, { "epoch": 0.79, "learning_rate": 8.983796759351872e-06, "loss": 0.294, "step": 2548 }, { "epoch": 0.79, "learning_rate": 8.983396679335869e-06, "loss": 0.3545, "step": 2549 }, { "epoch": 0.79, "learning_rate": 8.982996599319864e-06, "loss": 0.2588, "step": 2550 }, { "epoch": 0.79, "learning_rate": 8.98259651930386e-06, "loss": 0.2799, "step": 2551 }, { "epoch": 0.79, "learning_rate": 8.982196439287859e-06, "loss": 0.1587, "step": 2552 }, { "epoch": 0.79, "learning_rate": 8.981796359271855e-06, "loss": 0.4933, "step": 2553 }, { "epoch": 0.79, "learning_rate": 8.981396279255852e-06, "loss": 0.2794, "step": 2554 }, { "epoch": 0.79, "learning_rate": 8.980996199239849e-06, "loss": 0.37, "step": 2555 }, { "epoch": 0.79, "learning_rate": 8.980596119223845e-06, "loss": 0.2597, "step": 2556 }, { "epoch": 0.79, "learning_rate": 8.980196039207842e-06, "loss": 0.2812, "step": 2557 }, { "epoch": 0.79, "learning_rate": 8.979795959191839e-06, "loss": 0.2385, "step": 2558 }, { "epoch": 0.79, "learning_rate": 8.979395879175836e-06, "loss": 0.3703, "step": 2559 }, { "epoch": 0.79, "learning_rate": 8.978995799159834e-06, "loss": 0.3279, "step": 2560 }, { "epoch": 0.79, "learning_rate": 8.97859571914383e-06, "loss": 0.2026, "step": 2561 }, { "epoch": 0.8, "learning_rate": 8.978195639127826e-06, "loss": 0.3265, "step": 2562 }, { "epoch": 0.8, "learning_rate": 8.977795559111822e-06, "loss": 0.2877, "step": 2563 }, { "epoch": 0.8, "learning_rate": 8.977395479095819e-06, "loss": 0.3752, "step": 2564 }, { "epoch": 0.8, "learning_rate": 8.976995399079816e-06, "loss": 0.2962, "step": 2565 }, { "epoch": 0.8, "learning_rate": 8.976595319063814e-06, "loss": 0.3673, "step": 2566 }, { "epoch": 0.8, "learning_rate": 8.97619523904781e-06, "loss": 0.3159, "step": 2567 }, { "epoch": 0.8, "learning_rate": 8.975795159031807e-06, "loss": 0.2673, "step": 2568 }, { "epoch": 0.8, "learning_rate": 8.975395079015804e-06, "loss": 0.2019, "step": 2569 }, { "epoch": 0.8, "learning_rate": 8.9749949989998e-06, "loss": 0.246, "step": 2570 }, { "epoch": 0.8, "learning_rate": 8.974594918983797e-06, "loss": 0.2734, "step": 2571 }, { "epoch": 0.8, "learning_rate": 8.974194838967794e-06, "loss": 0.272, "step": 2572 }, { "epoch": 0.8, "learning_rate": 8.97379475895179e-06, "loss": 0.1749, "step": 2573 }, { "epoch": 0.8, "learning_rate": 8.97339467893579e-06, "loss": 0.3484, "step": 2574 }, { "epoch": 0.8, "learning_rate": 8.972994598919784e-06, "loss": 0.2298, "step": 2575 }, { "epoch": 0.8, "learning_rate": 8.972594518903781e-06, "loss": 0.1629, "step": 2576 }, { "epoch": 0.8, "learning_rate": 8.972194438887778e-06, "loss": 0.2407, "step": 2577 }, { "epoch": 0.8, "learning_rate": 8.971794358871774e-06, "loss": 0.3328, "step": 2578 }, { "epoch": 0.8, "learning_rate": 8.971394278855771e-06, "loss": 0.187, "step": 2579 }, { "epoch": 0.8, "learning_rate": 8.97099419883977e-06, "loss": 0.3947, "step": 2580 }, { "epoch": 0.8, "learning_rate": 8.970594118823766e-06, "loss": 0.2368, "step": 2581 }, { "epoch": 0.8, "learning_rate": 8.970194038807763e-06, "loss": 0.153, "step": 2582 }, { "epoch": 0.8, "learning_rate": 8.96979395879176e-06, "loss": 0.2701, "step": 2583 }, { "epoch": 0.8, "learning_rate": 8.969393878775756e-06, "loss": 0.334, "step": 2584 }, { "epoch": 0.8, "learning_rate": 8.968993798759753e-06, "loss": 0.2351, "step": 2585 }, { "epoch": 0.8, "learning_rate": 8.96859371874375e-06, "loss": 0.3221, "step": 2586 }, { "epoch": 0.8, "learning_rate": 8.968193638727746e-06, "loss": 0.1799, "step": 2587 }, { "epoch": 0.8, "learning_rate": 8.967793558711743e-06, "loss": 0.3306, "step": 2588 }, { "epoch": 0.8, "learning_rate": 8.96739347869574e-06, "loss": 0.3646, "step": 2589 }, { "epoch": 0.8, "learning_rate": 8.966993398679736e-06, "loss": 0.219, "step": 2590 }, { "epoch": 0.8, "learning_rate": 8.966593318663733e-06, "loss": 0.3519, "step": 2591 }, { "epoch": 0.8, "learning_rate": 8.96619323864773e-06, "loss": 0.3439, "step": 2592 }, { "epoch": 0.8, "learning_rate": 8.965793158631726e-06, "loss": 0.206, "step": 2593 }, { "epoch": 0.81, "learning_rate": 8.965393078615725e-06, "loss": 0.2225, "step": 2594 }, { "epoch": 0.81, "learning_rate": 8.964992998599721e-06, "loss": 0.1914, "step": 2595 }, { "epoch": 0.81, "learning_rate": 8.964592918583718e-06, "loss": 0.2229, "step": 2596 }, { "epoch": 0.81, "learning_rate": 8.964192838567715e-06, "loss": 0.3246, "step": 2597 }, { "epoch": 0.81, "learning_rate": 8.963792758551711e-06, "loss": 0.2814, "step": 2598 }, { "epoch": 0.81, "learning_rate": 8.963392678535708e-06, "loss": 0.1856, "step": 2599 }, { "epoch": 0.81, "learning_rate": 8.962992598519705e-06, "loss": 0.1871, "step": 2600 }, { "epoch": 0.81, "learning_rate": 8.962592518503701e-06, "loss": 0.3552, "step": 2601 }, { "epoch": 0.81, "learning_rate": 8.962192438487698e-06, "loss": 0.303, "step": 2602 }, { "epoch": 0.81, "learning_rate": 8.961792358471695e-06, "loss": 0.1673, "step": 2603 }, { "epoch": 0.81, "learning_rate": 8.961392278455691e-06, "loss": 0.2264, "step": 2604 }, { "epoch": 0.81, "learning_rate": 8.960992198439688e-06, "loss": 0.3473, "step": 2605 }, { "epoch": 0.81, "learning_rate": 8.960592118423685e-06, "loss": 0.2482, "step": 2606 }, { "epoch": 0.81, "learning_rate": 8.960192038407682e-06, "loss": 0.2056, "step": 2607 }, { "epoch": 0.81, "learning_rate": 8.95979195839168e-06, "loss": 0.1947, "step": 2608 }, { "epoch": 0.81, "learning_rate": 8.959391878375677e-06, "loss": 0.3182, "step": 2609 }, { "epoch": 0.81, "learning_rate": 8.958991798359673e-06, "loss": 0.2854, "step": 2610 }, { "epoch": 0.81, "learning_rate": 8.95859171834367e-06, "loss": 0.2704, "step": 2611 }, { "epoch": 0.81, "learning_rate": 8.958191638327667e-06, "loss": 0.2717, "step": 2612 }, { "epoch": 0.81, "learning_rate": 8.957791558311663e-06, "loss": 0.2812, "step": 2613 }, { "epoch": 0.81, "learning_rate": 8.95739147829566e-06, "loss": 0.1914, "step": 2614 }, { "epoch": 0.81, "learning_rate": 8.956991398279657e-06, "loss": 0.2748, "step": 2615 }, { "epoch": 0.81, "learning_rate": 8.956591318263653e-06, "loss": 0.2303, "step": 2616 }, { "epoch": 0.81, "learning_rate": 8.95619123824765e-06, "loss": 0.3586, "step": 2617 }, { "epoch": 0.81, "learning_rate": 8.955791158231647e-06, "loss": 0.2733, "step": 2618 }, { "epoch": 0.81, "learning_rate": 8.955391078215643e-06, "loss": 0.1583, "step": 2619 }, { "epoch": 0.81, "learning_rate": 8.95499099819964e-06, "loss": 0.1232, "step": 2620 }, { "epoch": 0.81, "learning_rate": 8.954590918183639e-06, "loss": 0.3632, "step": 2621 }, { "epoch": 0.81, "learning_rate": 8.954190838167635e-06, "loss": 0.226, "step": 2622 }, { "epoch": 0.81, "learning_rate": 8.953790758151632e-06, "loss": 0.1574, "step": 2623 }, { "epoch": 0.81, "learning_rate": 8.953390678135629e-06, "loss": 0.3218, "step": 2624 }, { "epoch": 0.81, "learning_rate": 8.952990598119624e-06, "loss": 0.2792, "step": 2625 }, { "epoch": 0.82, "learning_rate": 8.95259051810362e-06, "loss": 0.2495, "step": 2626 }, { "epoch": 0.82, "learning_rate": 8.952190438087619e-06, "loss": 0.232, "step": 2627 }, { "epoch": 0.82, "learning_rate": 8.951790358071615e-06, "loss": 0.25, "step": 2628 }, { "epoch": 0.82, "learning_rate": 8.951390278055612e-06, "loss": 0.1451, "step": 2629 }, { "epoch": 0.82, "learning_rate": 8.950990198039609e-06, "loss": 0.2575, "step": 2630 }, { "epoch": 0.82, "learning_rate": 8.950590118023605e-06, "loss": 0.1623, "step": 2631 }, { "epoch": 0.82, "learning_rate": 8.950190038007602e-06, "loss": 0.1991, "step": 2632 }, { "epoch": 0.82, "learning_rate": 8.949789957991599e-06, "loss": 0.3704, "step": 2633 }, { "epoch": 0.82, "learning_rate": 8.949389877975595e-06, "loss": 0.3822, "step": 2634 }, { "epoch": 0.82, "learning_rate": 8.948989797959594e-06, "loss": 0.2766, "step": 2635 }, { "epoch": 0.82, "learning_rate": 8.94858971794359e-06, "loss": 0.204, "step": 2636 }, { "epoch": 0.82, "learning_rate": 8.948189637927587e-06, "loss": 0.1688, "step": 2637 }, { "epoch": 0.82, "learning_rate": 8.947789557911582e-06, "loss": 0.2943, "step": 2638 }, { "epoch": 0.82, "learning_rate": 8.947389477895579e-06, "loss": 0.1859, "step": 2639 }, { "epoch": 0.82, "learning_rate": 8.946989397879576e-06, "loss": 0.3038, "step": 2640 }, { "epoch": 0.82, "learning_rate": 8.946589317863574e-06, "loss": 0.2698, "step": 2641 }, { "epoch": 0.82, "learning_rate": 8.94618923784757e-06, "loss": 0.2384, "step": 2642 }, { "epoch": 0.82, "learning_rate": 8.945789157831567e-06, "loss": 0.4322, "step": 2643 }, { "epoch": 0.82, "learning_rate": 8.945389077815564e-06, "loss": 0.4264, "step": 2644 }, { "epoch": 0.82, "learning_rate": 8.94498899779956e-06, "loss": 0.2421, "step": 2645 }, { "epoch": 0.82, "learning_rate": 8.944588917783557e-06, "loss": 0.3134, "step": 2646 }, { "epoch": 0.82, "learning_rate": 8.944188837767554e-06, "loss": 0.234, "step": 2647 }, { "epoch": 0.82, "learning_rate": 8.94378875775155e-06, "loss": 0.2369, "step": 2648 }, { "epoch": 0.82, "learning_rate": 8.943388677735549e-06, "loss": 0.3147, "step": 2649 }, { "epoch": 0.82, "learning_rate": 8.942988597719546e-06, "loss": 0.2918, "step": 2650 }, { "epoch": 0.82, "learning_rate": 8.94258851770354e-06, "loss": 0.3063, "step": 2651 }, { "epoch": 0.82, "learning_rate": 8.942188437687537e-06, "loss": 0.2577, "step": 2652 }, { "epoch": 0.82, "learning_rate": 8.941788357671534e-06, "loss": 0.2948, "step": 2653 }, { "epoch": 0.82, "learning_rate": 8.941388277655531e-06, "loss": 0.3122, "step": 2654 }, { "epoch": 0.82, "learning_rate": 8.94098819763953e-06, "loss": 0.3038, "step": 2655 }, { "epoch": 0.82, "learning_rate": 8.940588117623526e-06, "loss": 0.2535, "step": 2656 }, { "epoch": 0.82, "learning_rate": 8.940188037607523e-06, "loss": 0.3011, "step": 2657 }, { "epoch": 0.82, "learning_rate": 8.93978795759152e-06, "loss": 0.2183, "step": 2658 }, { "epoch": 0.83, "learning_rate": 8.939387877575516e-06, "loss": 0.2497, "step": 2659 }, { "epoch": 0.83, "learning_rate": 8.938987797559513e-06, "loss": 0.2479, "step": 2660 }, { "epoch": 0.83, "learning_rate": 8.93858771754351e-06, "loss": 0.3889, "step": 2661 }, { "epoch": 0.83, "learning_rate": 8.938187637527506e-06, "loss": 0.1868, "step": 2662 }, { "epoch": 0.83, "learning_rate": 8.937787557511503e-06, "loss": 0.2224, "step": 2663 }, { "epoch": 0.83, "learning_rate": 8.9373874774955e-06, "loss": 0.2309, "step": 2664 }, { "epoch": 0.83, "learning_rate": 8.936987397479496e-06, "loss": 0.2198, "step": 2665 }, { "epoch": 0.83, "learning_rate": 8.936587317463493e-06, "loss": 0.2712, "step": 2666 }, { "epoch": 0.83, "learning_rate": 8.93618723744749e-06, "loss": 0.3185, "step": 2667 }, { "epoch": 0.83, "learning_rate": 8.935787157431486e-06, "loss": 0.3044, "step": 2668 }, { "epoch": 0.83, "learning_rate": 8.935387077415485e-06, "loss": 0.2338, "step": 2669 }, { "epoch": 0.83, "learning_rate": 8.934986997399481e-06, "loss": 0.229, "step": 2670 }, { "epoch": 0.83, "learning_rate": 8.934586917383478e-06, "loss": 0.2534, "step": 2671 }, { "epoch": 0.83, "learning_rate": 8.934186837367475e-06, "loss": 0.1766, "step": 2672 }, { "epoch": 0.83, "learning_rate": 8.933786757351471e-06, "loss": 0.357, "step": 2673 }, { "epoch": 0.83, "learning_rate": 8.933386677335468e-06, "loss": 0.4494, "step": 2674 }, { "epoch": 0.83, "learning_rate": 8.932986597319465e-06, "loss": 0.2719, "step": 2675 }, { "epoch": 0.83, "learning_rate": 8.932586517303461e-06, "loss": 0.3273, "step": 2676 }, { "epoch": 0.83, "learning_rate": 8.932186437287458e-06, "loss": 0.2705, "step": 2677 }, { "epoch": 0.83, "learning_rate": 8.931786357271455e-06, "loss": 0.342, "step": 2678 }, { "epoch": 0.83, "learning_rate": 8.931386277255451e-06, "loss": 0.2884, "step": 2679 }, { "epoch": 0.83, "learning_rate": 8.930986197239448e-06, "loss": 0.2202, "step": 2680 }, { "epoch": 0.83, "learning_rate": 8.930586117223445e-06, "loss": 0.1615, "step": 2681 }, { "epoch": 0.83, "learning_rate": 8.930186037207441e-06, "loss": 0.3573, "step": 2682 }, { "epoch": 0.83, "learning_rate": 8.92978595719144e-06, "loss": 0.2371, "step": 2683 }, { "epoch": 0.83, "learning_rate": 8.929385877175437e-06, "loss": 0.2036, "step": 2684 }, { "epoch": 0.83, "learning_rate": 8.928985797159433e-06, "loss": 0.2841, "step": 2685 }, { "epoch": 0.83, "learning_rate": 8.92858571714343e-06, "loss": 0.1643, "step": 2686 }, { "epoch": 0.83, "learning_rate": 8.928185637127427e-06, "loss": 0.1271, "step": 2687 }, { "epoch": 0.83, "learning_rate": 8.927785557111423e-06, "loss": 0.2882, "step": 2688 }, { "epoch": 0.83, "learning_rate": 8.92738547709542e-06, "loss": 0.3065, "step": 2689 }, { "epoch": 0.83, "learning_rate": 8.926985397079417e-06, "loss": 0.2154, "step": 2690 }, { "epoch": 0.84, "learning_rate": 8.926585317063413e-06, "loss": 0.2973, "step": 2691 }, { "epoch": 0.84, "learning_rate": 8.92618523704741e-06, "loss": 0.2994, "step": 2692 }, { "epoch": 0.84, "learning_rate": 8.925785157031407e-06, "loss": 0.2106, "step": 2693 }, { "epoch": 0.84, "learning_rate": 8.925385077015403e-06, "loss": 0.2263, "step": 2694 }, { "epoch": 0.84, "learning_rate": 8.9249849969994e-06, "loss": 0.2435, "step": 2695 }, { "epoch": 0.84, "learning_rate": 8.924584916983397e-06, "loss": 0.1816, "step": 2696 }, { "epoch": 0.84, "learning_rate": 8.924184836967395e-06, "loss": 0.2484, "step": 2697 }, { "epoch": 0.84, "learning_rate": 8.923784756951392e-06, "loss": 0.2242, "step": 2698 }, { "epoch": 0.84, "learning_rate": 8.923384676935389e-06, "loss": 0.3632, "step": 2699 }, { "epoch": 0.84, "learning_rate": 8.922984596919385e-06, "loss": 0.3008, "step": 2700 }, { "epoch": 0.84, "learning_rate": 8.92258451690338e-06, "loss": 0.5108, "step": 2701 }, { "epoch": 0.84, "learning_rate": 8.922184436887379e-06, "loss": 0.2655, "step": 2702 }, { "epoch": 0.84, "learning_rate": 8.921784356871375e-06, "loss": 0.2031, "step": 2703 }, { "epoch": 0.84, "learning_rate": 8.921384276855372e-06, "loss": 0.2781, "step": 2704 }, { "epoch": 0.84, "learning_rate": 8.920984196839369e-06, "loss": 0.335, "step": 2705 }, { "epoch": 0.84, "learning_rate": 8.920584116823365e-06, "loss": 0.0814, "step": 2706 }, { "epoch": 0.84, "learning_rate": 8.920184036807362e-06, "loss": 0.1694, "step": 2707 }, { "epoch": 0.84, "learning_rate": 8.919783956791359e-06, "loss": 0.2528, "step": 2708 }, { "epoch": 0.84, "learning_rate": 8.919383876775355e-06, "loss": 0.3262, "step": 2709 }, { "epoch": 0.84, "learning_rate": 8.918983796759352e-06, "loss": 0.3228, "step": 2710 }, { "epoch": 0.84, "learning_rate": 8.91858371674335e-06, "loss": 0.2511, "step": 2711 }, { "epoch": 0.84, "learning_rate": 8.918183636727347e-06, "loss": 0.2946, "step": 2712 }, { "epoch": 0.84, "learning_rate": 8.917783556711344e-06, "loss": 0.2697, "step": 2713 }, { "epoch": 0.84, "learning_rate": 8.917383476695339e-06, "loss": 0.1913, "step": 2714 }, { "epoch": 0.84, "learning_rate": 8.916983396679336e-06, "loss": 0.2675, "step": 2715 }, { "epoch": 0.84, "learning_rate": 8.916583316663334e-06, "loss": 0.26, "step": 2716 }, { "epoch": 0.84, "learning_rate": 8.91618323664733e-06, "loss": 0.3013, "step": 2717 }, { "epoch": 0.84, "learning_rate": 8.915783156631327e-06, "loss": 0.2419, "step": 2718 }, { "epoch": 0.84, "learning_rate": 8.915383076615324e-06, "loss": 0.224, "step": 2719 }, { "epoch": 0.84, "learning_rate": 8.91498299659932e-06, "loss": 0.2417, "step": 2720 }, { "epoch": 0.84, "learning_rate": 8.914582916583317e-06, "loss": 0.1886, "step": 2721 }, { "epoch": 0.84, "learning_rate": 8.914182836567314e-06, "loss": 0.2694, "step": 2722 }, { "epoch": 0.85, "learning_rate": 8.91378275655131e-06, "loss": 0.2937, "step": 2723 }, { "epoch": 0.85, "learning_rate": 8.913382676535307e-06, "loss": 0.2086, "step": 2724 }, { "epoch": 0.85, "learning_rate": 8.912982596519306e-06, "loss": 0.3123, "step": 2725 }, { "epoch": 0.85, "learning_rate": 8.9125825165033e-06, "loss": 0.3842, "step": 2726 }, { "epoch": 0.85, "learning_rate": 8.912182436487297e-06, "loss": 0.2075, "step": 2727 }, { "epoch": 0.85, "learning_rate": 8.911782356471294e-06, "loss": 0.3178, "step": 2728 }, { "epoch": 0.85, "learning_rate": 8.91138227645529e-06, "loss": 0.1399, "step": 2729 }, { "epoch": 0.85, "learning_rate": 8.91098219643929e-06, "loss": 0.2389, "step": 2730 }, { "epoch": 0.85, "learning_rate": 8.910582116423286e-06, "loss": 0.2501, "step": 2731 }, { "epoch": 0.85, "learning_rate": 8.910182036407283e-06, "loss": 0.3567, "step": 2732 }, { "epoch": 0.85, "learning_rate": 8.90978195639128e-06, "loss": 0.3777, "step": 2733 }, { "epoch": 0.85, "learning_rate": 8.909381876375276e-06, "loss": 0.3513, "step": 2734 }, { "epoch": 0.85, "learning_rate": 8.908981796359273e-06, "loss": 0.2186, "step": 2735 }, { "epoch": 0.85, "learning_rate": 8.90858171634327e-06, "loss": 0.1646, "step": 2736 }, { "epoch": 0.85, "learning_rate": 8.908181636327266e-06, "loss": 0.3517, "step": 2737 }, { "epoch": 0.85, "learning_rate": 8.907781556311263e-06, "loss": 0.2282, "step": 2738 }, { "epoch": 0.85, "learning_rate": 8.90738147629526e-06, "loss": 0.1075, "step": 2739 }, { "epoch": 0.85, "learning_rate": 8.906981396279256e-06, "loss": 0.1116, "step": 2740 }, { "epoch": 0.85, "learning_rate": 8.906581316263253e-06, "loss": 0.2407, "step": 2741 }, { "epoch": 0.85, "learning_rate": 8.90618123624725e-06, "loss": 0.3743, "step": 2742 }, { "epoch": 0.85, "learning_rate": 8.905781156231246e-06, "loss": 0.1914, "step": 2743 }, { "epoch": 0.85, "learning_rate": 8.905381076215244e-06, "loss": 0.237, "step": 2744 }, { "epoch": 0.85, "learning_rate": 8.904980996199241e-06, "loss": 0.2514, "step": 2745 }, { "epoch": 0.85, "learning_rate": 8.904580916183238e-06, "loss": 0.3679, "step": 2746 }, { "epoch": 0.85, "learning_rate": 8.904180836167235e-06, "loss": 0.2289, "step": 2747 }, { "epoch": 0.85, "learning_rate": 8.903780756151231e-06, "loss": 0.4087, "step": 2748 }, { "epoch": 0.85, "learning_rate": 8.903380676135228e-06, "loss": 0.1856, "step": 2749 }, { "epoch": 0.85, "learning_rate": 8.902980596119225e-06, "loss": 0.2119, "step": 2750 }, { "epoch": 0.85, "learning_rate": 8.902580516103221e-06, "loss": 0.2668, "step": 2751 }, { "epoch": 0.85, "learning_rate": 8.902180436087218e-06, "loss": 0.3306, "step": 2752 }, { "epoch": 0.85, "learning_rate": 8.901780356071215e-06, "loss": 0.1734, "step": 2753 }, { "epoch": 0.85, "learning_rate": 8.901380276055211e-06, "loss": 0.1571, "step": 2754 }, { "epoch": 0.86, "learning_rate": 8.900980196039208e-06, "loss": 0.2158, "step": 2755 }, { "epoch": 0.86, "learning_rate": 8.900580116023205e-06, "loss": 0.4078, "step": 2756 }, { "epoch": 0.86, "learning_rate": 8.900180036007201e-06, "loss": 0.406, "step": 2757 }, { "epoch": 0.86, "learning_rate": 8.8997799559912e-06, "loss": 0.2688, "step": 2758 }, { "epoch": 0.86, "learning_rate": 8.899379875975196e-06, "loss": 0.1372, "step": 2759 }, { "epoch": 0.86, "learning_rate": 8.898979795959193e-06, "loss": 0.31, "step": 2760 }, { "epoch": 0.86, "learning_rate": 8.89857971594319e-06, "loss": 0.1246, "step": 2761 }, { "epoch": 0.86, "learning_rate": 8.898179635927187e-06, "loss": 0.2431, "step": 2762 }, { "epoch": 0.86, "learning_rate": 8.897779555911183e-06, "loss": 0.1812, "step": 2763 }, { "epoch": 0.86, "learning_rate": 8.89737947589518e-06, "loss": 0.2219, "step": 2764 }, { "epoch": 0.86, "learning_rate": 8.896979395879177e-06, "loss": 0.1428, "step": 2765 }, { "epoch": 0.86, "learning_rate": 8.896579315863173e-06, "loss": 0.1743, "step": 2766 }, { "epoch": 0.86, "learning_rate": 8.89617923584717e-06, "loss": 0.3023, "step": 2767 }, { "epoch": 0.86, "learning_rate": 8.895779155831167e-06, "loss": 0.2257, "step": 2768 }, { "epoch": 0.86, "learning_rate": 8.895379075815163e-06, "loss": 0.3114, "step": 2769 }, { "epoch": 0.86, "learning_rate": 8.89497899579916e-06, "loss": 0.2325, "step": 2770 }, { "epoch": 0.86, "learning_rate": 8.894578915783157e-06, "loss": 0.2152, "step": 2771 }, { "epoch": 0.86, "learning_rate": 8.894178835767155e-06, "loss": 0.1721, "step": 2772 }, { "epoch": 0.86, "learning_rate": 8.893778755751152e-06, "loss": 0.1337, "step": 2773 }, { "epoch": 0.86, "learning_rate": 8.893378675735148e-06, "loss": 0.2805, "step": 2774 }, { "epoch": 0.86, "learning_rate": 8.892978595719145e-06, "loss": 0.1705, "step": 2775 }, { "epoch": 0.86, "learning_rate": 8.892578515703142e-06, "loss": 0.1778, "step": 2776 }, { "epoch": 0.86, "learning_rate": 8.892178435687139e-06, "loss": 0.1604, "step": 2777 }, { "epoch": 0.86, "learning_rate": 8.891778355671135e-06, "loss": 0.2095, "step": 2778 }, { "epoch": 0.86, "learning_rate": 8.891378275655132e-06, "loss": 0.1983, "step": 2779 }, { "epoch": 0.86, "learning_rate": 8.890978195639129e-06, "loss": 0.1574, "step": 2780 }, { "epoch": 0.86, "learning_rate": 8.890578115623125e-06, "loss": 0.374, "step": 2781 }, { "epoch": 0.86, "learning_rate": 8.890178035607122e-06, "loss": 0.2104, "step": 2782 }, { "epoch": 0.86, "learning_rate": 8.889777955591119e-06, "loss": 0.2492, "step": 2783 }, { "epoch": 0.86, "learning_rate": 8.889377875575115e-06, "loss": 0.2532, "step": 2784 }, { "epoch": 0.86, "learning_rate": 8.888977795559112e-06, "loss": 0.218, "step": 2785 }, { "epoch": 0.86, "learning_rate": 8.88857771554311e-06, "loss": 0.2375, "step": 2786 }, { "epoch": 0.87, "learning_rate": 8.888177635527107e-06, "loss": 0.2561, "step": 2787 }, { "epoch": 0.87, "learning_rate": 8.887777555511104e-06, "loss": 0.3115, "step": 2788 }, { "epoch": 0.87, "learning_rate": 8.887377475495099e-06, "loss": 0.3424, "step": 2789 }, { "epoch": 0.87, "learning_rate": 8.886977395479095e-06, "loss": 0.1756, "step": 2790 }, { "epoch": 0.87, "learning_rate": 8.886577315463094e-06, "loss": 0.245, "step": 2791 }, { "epoch": 0.87, "learning_rate": 8.88617723544709e-06, "loss": 0.2468, "step": 2792 }, { "epoch": 0.87, "learning_rate": 8.885777155431087e-06, "loss": 0.1699, "step": 2793 }, { "epoch": 0.87, "learning_rate": 8.885377075415084e-06, "loss": 0.3039, "step": 2794 }, { "epoch": 0.87, "learning_rate": 8.88497699539908e-06, "loss": 0.2465, "step": 2795 }, { "epoch": 0.87, "learning_rate": 8.884576915383077e-06, "loss": 0.2787, "step": 2796 }, { "epoch": 0.87, "learning_rate": 8.884176835367074e-06, "loss": 0.1529, "step": 2797 }, { "epoch": 0.87, "learning_rate": 8.88377675535107e-06, "loss": 0.1731, "step": 2798 }, { "epoch": 0.87, "learning_rate": 8.883376675335067e-06, "loss": 0.2745, "step": 2799 }, { "epoch": 0.87, "learning_rate": 8.882976595319066e-06, "loss": 0.3878, "step": 2800 }, { "epoch": 0.87, "learning_rate": 8.882576515303062e-06, "loss": 0.2772, "step": 2801 }, { "epoch": 0.87, "learning_rate": 8.882176435287057e-06, "loss": 0.2672, "step": 2802 }, { "epoch": 0.87, "learning_rate": 8.881776355271054e-06, "loss": 0.2804, "step": 2803 }, { "epoch": 0.87, "learning_rate": 8.88137627525505e-06, "loss": 0.2721, "step": 2804 }, { "epoch": 0.87, "learning_rate": 8.880976195239049e-06, "loss": 0.1281, "step": 2805 }, { "epoch": 0.87, "learning_rate": 8.880576115223046e-06, "loss": 0.202, "step": 2806 }, { "epoch": 0.87, "learning_rate": 8.880176035207042e-06, "loss": 0.1636, "step": 2807 }, { "epoch": 0.87, "learning_rate": 8.87977595519104e-06, "loss": 0.1628, "step": 2808 }, { "epoch": 0.87, "learning_rate": 8.879375875175036e-06, "loss": 0.1475, "step": 2809 }, { "epoch": 0.87, "learning_rate": 8.878975795159033e-06, "loss": 0.2865, "step": 2810 }, { "epoch": 0.87, "learning_rate": 8.87857571514303e-06, "loss": 0.2707, "step": 2811 }, { "epoch": 0.87, "learning_rate": 8.878175635127026e-06, "loss": 0.1426, "step": 2812 }, { "epoch": 0.87, "learning_rate": 8.877775555111023e-06, "loss": 0.3145, "step": 2813 }, { "epoch": 0.87, "learning_rate": 8.87737547509502e-06, "loss": 0.2669, "step": 2814 }, { "epoch": 0.87, "learning_rate": 8.876975395079016e-06, "loss": 0.2392, "step": 2815 }, { "epoch": 0.87, "learning_rate": 8.876575315063013e-06, "loss": 0.2844, "step": 2816 }, { "epoch": 0.87, "learning_rate": 8.87617523504701e-06, "loss": 0.2437, "step": 2817 }, { "epoch": 0.87, "learning_rate": 8.875775155031006e-06, "loss": 0.2694, "step": 2818 }, { "epoch": 0.87, "learning_rate": 8.875375075015004e-06, "loss": 0.2592, "step": 2819 }, { "epoch": 0.88, "learning_rate": 8.874974994999001e-06, "loss": 0.0756, "step": 2820 }, { "epoch": 0.88, "learning_rate": 8.874574914982998e-06, "loss": 0.198, "step": 2821 }, { "epoch": 0.88, "learning_rate": 8.874174834966994e-06, "loss": 0.3985, "step": 2822 }, { "epoch": 0.88, "learning_rate": 8.873774754950991e-06, "loss": 0.3146, "step": 2823 }, { "epoch": 0.88, "learning_rate": 8.873374674934988e-06, "loss": 0.2905, "step": 2824 }, { "epoch": 0.88, "learning_rate": 8.872974594918985e-06, "loss": 0.1138, "step": 2825 }, { "epoch": 0.88, "learning_rate": 8.872574514902981e-06, "loss": 0.2519, "step": 2826 }, { "epoch": 0.88, "learning_rate": 8.872174434886978e-06, "loss": 0.4113, "step": 2827 }, { "epoch": 0.88, "learning_rate": 8.871774354870975e-06, "loss": 0.1958, "step": 2828 }, { "epoch": 0.88, "learning_rate": 8.871374274854971e-06, "loss": 0.2079, "step": 2829 }, { "epoch": 0.88, "learning_rate": 8.870974194838968e-06, "loss": 0.489, "step": 2830 }, { "epoch": 0.88, "learning_rate": 8.870574114822965e-06, "loss": 0.3241, "step": 2831 }, { "epoch": 0.88, "learning_rate": 8.870174034806961e-06, "loss": 0.1764, "step": 2832 }, { "epoch": 0.88, "learning_rate": 8.86977395479096e-06, "loss": 0.1982, "step": 2833 }, { "epoch": 0.88, "learning_rate": 8.869373874774956e-06, "loss": 0.2603, "step": 2834 }, { "epoch": 0.88, "learning_rate": 8.868973794758953e-06, "loss": 0.3305, "step": 2835 }, { "epoch": 0.88, "learning_rate": 8.86857371474295e-06, "loss": 0.2339, "step": 2836 }, { "epoch": 0.88, "learning_rate": 8.868173634726946e-06, "loss": 0.2501, "step": 2837 }, { "epoch": 0.88, "learning_rate": 8.867773554710943e-06, "loss": 0.3557, "step": 2838 }, { "epoch": 0.88, "learning_rate": 8.86737347469494e-06, "loss": 0.3533, "step": 2839 }, { "epoch": 0.88, "learning_rate": 8.866973394678937e-06, "loss": 0.1415, "step": 2840 }, { "epoch": 0.88, "learning_rate": 8.866573314662933e-06, "loss": 0.2533, "step": 2841 }, { "epoch": 0.88, "learning_rate": 8.86617323464693e-06, "loss": 0.3082, "step": 2842 }, { "epoch": 0.88, "learning_rate": 8.865773154630927e-06, "loss": 0.251, "step": 2843 }, { "epoch": 0.88, "learning_rate": 8.865373074614923e-06, "loss": 0.2141, "step": 2844 }, { "epoch": 0.88, "learning_rate": 8.86497299459892e-06, "loss": 0.2186, "step": 2845 }, { "epoch": 0.88, "learning_rate": 8.864572914582917e-06, "loss": 0.3946, "step": 2846 }, { "epoch": 0.88, "learning_rate": 8.864172834566915e-06, "loss": 0.295, "step": 2847 }, { "epoch": 0.88, "learning_rate": 8.863772754550912e-06, "loss": 0.3116, "step": 2848 }, { "epoch": 0.88, "learning_rate": 8.863372674534908e-06, "loss": 0.1493, "step": 2849 }, { "epoch": 0.88, "learning_rate": 8.862972594518905e-06, "loss": 0.1587, "step": 2850 }, { "epoch": 0.88, "learning_rate": 8.862572514502902e-06, "loss": 0.3242, "step": 2851 }, { "epoch": 0.89, "learning_rate": 8.862172434486897e-06, "loss": 0.3226, "step": 2852 }, { "epoch": 0.89, "learning_rate": 8.861772354470895e-06, "loss": 0.2557, "step": 2853 }, { "epoch": 0.89, "learning_rate": 8.861372274454892e-06, "loss": 0.1952, "step": 2854 }, { "epoch": 0.89, "learning_rate": 8.860972194438889e-06, "loss": 0.3255, "step": 2855 }, { "epoch": 0.89, "learning_rate": 8.860572114422885e-06, "loss": 0.2113, "step": 2856 }, { "epoch": 0.89, "learning_rate": 8.860172034406882e-06, "loss": 0.3306, "step": 2857 }, { "epoch": 0.89, "learning_rate": 8.859771954390879e-06, "loss": 0.2234, "step": 2858 }, { "epoch": 0.89, "learning_rate": 8.859371874374875e-06, "loss": 0.2793, "step": 2859 }, { "epoch": 0.89, "learning_rate": 8.858971794358872e-06, "loss": 0.2162, "step": 2860 }, { "epoch": 0.89, "learning_rate": 8.85857171434287e-06, "loss": 0.2134, "step": 2861 }, { "epoch": 0.89, "learning_rate": 8.858171634326867e-06, "loss": 0.2718, "step": 2862 }, { "epoch": 0.89, "learning_rate": 8.857771554310864e-06, "loss": 0.3813, "step": 2863 }, { "epoch": 0.89, "learning_rate": 8.85737147429486e-06, "loss": 0.2089, "step": 2864 }, { "epoch": 0.89, "learning_rate": 8.856971394278855e-06, "loss": 0.2707, "step": 2865 }, { "epoch": 0.89, "learning_rate": 8.856571314262852e-06, "loss": 0.1964, "step": 2866 }, { "epoch": 0.89, "learning_rate": 8.85617123424685e-06, "loss": 0.4424, "step": 2867 }, { "epoch": 0.89, "learning_rate": 8.855771154230847e-06, "loss": 0.2609, "step": 2868 }, { "epoch": 0.89, "learning_rate": 8.855371074214844e-06, "loss": 0.262, "step": 2869 }, { "epoch": 0.89, "learning_rate": 8.85497099419884e-06, "loss": 0.2109, "step": 2870 }, { "epoch": 0.89, "learning_rate": 8.854570914182837e-06, "loss": 0.1343, "step": 2871 }, { "epoch": 0.89, "learning_rate": 8.854170834166834e-06, "loss": 0.1877, "step": 2872 }, { "epoch": 0.89, "learning_rate": 8.85377075415083e-06, "loss": 0.4564, "step": 2873 }, { "epoch": 0.89, "learning_rate": 8.853370674134827e-06, "loss": 0.3081, "step": 2874 }, { "epoch": 0.89, "learning_rate": 8.852970594118826e-06, "loss": 0.2259, "step": 2875 }, { "epoch": 0.89, "learning_rate": 8.852570514102822e-06, "loss": 0.456, "step": 2876 }, { "epoch": 0.89, "learning_rate": 8.852170434086819e-06, "loss": 0.2888, "step": 2877 }, { "epoch": 0.89, "learning_rate": 8.851770354070814e-06, "loss": 0.2193, "step": 2878 }, { "epoch": 0.89, "learning_rate": 8.85137027405481e-06, "loss": 0.1701, "step": 2879 }, { "epoch": 0.89, "learning_rate": 8.850970194038807e-06, "loss": 0.2043, "step": 2880 }, { "epoch": 0.89, "learning_rate": 8.850570114022806e-06, "loss": 0.322, "step": 2881 }, { "epoch": 0.89, "learning_rate": 8.850170034006802e-06, "loss": 0.2664, "step": 2882 }, { "epoch": 0.89, "learning_rate": 8.849769953990799e-06, "loss": 0.2396, "step": 2883 }, { "epoch": 0.9, "learning_rate": 8.849369873974796e-06, "loss": 0.2282, "step": 2884 }, { "epoch": 0.9, "learning_rate": 8.848969793958792e-06, "loss": 0.1971, "step": 2885 }, { "epoch": 0.9, "learning_rate": 8.84856971394279e-06, "loss": 0.2232, "step": 2886 }, { "epoch": 0.9, "learning_rate": 8.848169633926786e-06, "loss": 0.2436, "step": 2887 }, { "epoch": 0.9, "learning_rate": 8.847769553910783e-06, "loss": 0.1617, "step": 2888 }, { "epoch": 0.9, "learning_rate": 8.847369473894781e-06, "loss": 0.287, "step": 2889 }, { "epoch": 0.9, "learning_rate": 8.846969393878776e-06, "loss": 0.2567, "step": 2890 }, { "epoch": 0.9, "learning_rate": 8.846569313862773e-06, "loss": 0.2773, "step": 2891 }, { "epoch": 0.9, "learning_rate": 8.84616923384677e-06, "loss": 0.2273, "step": 2892 }, { "epoch": 0.9, "learning_rate": 8.845769153830766e-06, "loss": 0.2298, "step": 2893 }, { "epoch": 0.9, "learning_rate": 8.845369073814763e-06, "loss": 0.1485, "step": 2894 }, { "epoch": 0.9, "learning_rate": 8.844968993798761e-06, "loss": 0.2898, "step": 2895 }, { "epoch": 0.9, "learning_rate": 8.844568913782758e-06, "loss": 0.2428, "step": 2896 }, { "epoch": 0.9, "learning_rate": 8.844168833766754e-06, "loss": 0.3605, "step": 2897 }, { "epoch": 0.9, "learning_rate": 8.843768753750751e-06, "loss": 0.2569, "step": 2898 }, { "epoch": 0.9, "learning_rate": 8.843368673734748e-06, "loss": 0.277, "step": 2899 }, { "epoch": 0.9, "learning_rate": 8.842968593718744e-06, "loss": 0.2999, "step": 2900 }, { "epoch": 0.9, "learning_rate": 8.842568513702741e-06, "loss": 0.237, "step": 2901 }, { "epoch": 0.9, "learning_rate": 8.842168433686738e-06, "loss": 0.2698, "step": 2902 }, { "epoch": 0.9, "learning_rate": 8.841768353670735e-06, "loss": 0.2961, "step": 2903 }, { "epoch": 0.9, "learning_rate": 8.841368273654731e-06, "loss": 0.234, "step": 2904 }, { "epoch": 0.9, "learning_rate": 8.840968193638728e-06, "loss": 0.237, "step": 2905 }, { "epoch": 0.9, "learning_rate": 8.840568113622725e-06, "loss": 0.1408, "step": 2906 }, { "epoch": 0.9, "learning_rate": 8.840168033606721e-06, "loss": 0.3005, "step": 2907 }, { "epoch": 0.9, "learning_rate": 8.839767953590718e-06, "loss": 0.1716, "step": 2908 }, { "epoch": 0.9, "learning_rate": 8.839367873574716e-06, "loss": 0.2364, "step": 2909 }, { "epoch": 0.9, "learning_rate": 8.838967793558713e-06, "loss": 0.2167, "step": 2910 }, { "epoch": 0.9, "learning_rate": 8.83856771354271e-06, "loss": 0.1917, "step": 2911 }, { "epoch": 0.9, "learning_rate": 8.838167633526706e-06, "loss": 0.365, "step": 2912 }, { "epoch": 0.9, "learning_rate": 8.837767553510703e-06, "loss": 0.1846, "step": 2913 }, { "epoch": 0.9, "learning_rate": 8.8373674734947e-06, "loss": 0.3158, "step": 2914 }, { "epoch": 0.9, "learning_rate": 8.836967393478696e-06, "loss": 0.4477, "step": 2915 }, { "epoch": 0.91, "learning_rate": 8.836567313462693e-06, "loss": 0.2412, "step": 2916 }, { "epoch": 0.91, "learning_rate": 8.83616723344669e-06, "loss": 0.2125, "step": 2917 }, { "epoch": 0.91, "learning_rate": 8.835767153430687e-06, "loss": 0.1921, "step": 2918 }, { "epoch": 0.91, "learning_rate": 8.835367073414683e-06, "loss": 0.2152, "step": 2919 }, { "epoch": 0.91, "learning_rate": 8.83496699339868e-06, "loss": 0.2341, "step": 2920 }, { "epoch": 0.91, "learning_rate": 8.834566913382677e-06, "loss": 0.2014, "step": 2921 }, { "epoch": 0.91, "learning_rate": 8.834166833366675e-06, "loss": 0.2635, "step": 2922 }, { "epoch": 0.91, "learning_rate": 8.833766753350672e-06, "loss": 0.5436, "step": 2923 }, { "epoch": 0.91, "learning_rate": 8.833366673334668e-06, "loss": 0.234, "step": 2924 }, { "epoch": 0.91, "learning_rate": 8.832966593318665e-06, "loss": 0.3083, "step": 2925 }, { "epoch": 0.91, "learning_rate": 8.832566513302662e-06, "loss": 0.3141, "step": 2926 }, { "epoch": 0.91, "learning_rate": 8.832166433286658e-06, "loss": 0.3323, "step": 2927 }, { "epoch": 0.91, "learning_rate": 8.831766353270655e-06, "loss": 0.3312, "step": 2928 }, { "epoch": 0.91, "learning_rate": 8.831366273254652e-06, "loss": 0.2984, "step": 2929 }, { "epoch": 0.91, "learning_rate": 8.830966193238648e-06, "loss": 0.1775, "step": 2930 }, { "epoch": 0.91, "learning_rate": 8.830566113222645e-06, "loss": 0.2487, "step": 2931 }, { "epoch": 0.91, "learning_rate": 8.830166033206642e-06, "loss": 0.1866, "step": 2932 }, { "epoch": 0.91, "learning_rate": 8.829765953190639e-06, "loss": 0.1343, "step": 2933 }, { "epoch": 0.91, "learning_rate": 8.829365873174635e-06, "loss": 0.2263, "step": 2934 }, { "epoch": 0.91, "learning_rate": 8.828965793158632e-06, "loss": 0.3055, "step": 2935 }, { "epoch": 0.91, "learning_rate": 8.82856571314263e-06, "loss": 0.2011, "step": 2936 }, { "epoch": 0.91, "learning_rate": 8.828165633126627e-06, "loss": 0.3215, "step": 2937 }, { "epoch": 0.91, "learning_rate": 8.827765553110624e-06, "loss": 0.2846, "step": 2938 }, { "epoch": 0.91, "learning_rate": 8.82736547309462e-06, "loss": 0.1934, "step": 2939 }, { "epoch": 0.91, "learning_rate": 8.826965393078617e-06, "loss": 0.1623, "step": 2940 }, { "epoch": 0.91, "learning_rate": 8.826565313062612e-06, "loss": 0.2384, "step": 2941 }, { "epoch": 0.91, "learning_rate": 8.82616523304661e-06, "loss": 0.2136, "step": 2942 }, { "epoch": 0.91, "learning_rate": 8.825765153030607e-06, "loss": 0.3226, "step": 2943 }, { "epoch": 0.91, "learning_rate": 8.825365073014604e-06, "loss": 0.2714, "step": 2944 }, { "epoch": 0.91, "learning_rate": 8.8249649929986e-06, "loss": 0.2137, "step": 2945 }, { "epoch": 0.91, "learning_rate": 8.824564912982597e-06, "loss": 0.2248, "step": 2946 }, { "epoch": 0.91, "learning_rate": 8.824164832966594e-06, "loss": 0.2396, "step": 2947 }, { "epoch": 0.91, "learning_rate": 8.82376475295059e-06, "loss": 0.2907, "step": 2948 }, { "epoch": 0.92, "learning_rate": 8.823364672934587e-06, "loss": 0.3052, "step": 2949 }, { "epoch": 0.92, "learning_rate": 8.822964592918586e-06, "loss": 0.2458, "step": 2950 }, { "epoch": 0.92, "learning_rate": 8.822564512902582e-06, "loss": 0.2219, "step": 2951 }, { "epoch": 0.92, "learning_rate": 8.822164432886579e-06, "loss": 0.2834, "step": 2952 }, { "epoch": 0.92, "learning_rate": 8.821764352870574e-06, "loss": 0.2946, "step": 2953 }, { "epoch": 0.92, "learning_rate": 8.82136427285457e-06, "loss": 0.2277, "step": 2954 }, { "epoch": 0.92, "learning_rate": 8.820964192838567e-06, "loss": 0.2319, "step": 2955 }, { "epoch": 0.92, "learning_rate": 8.820564112822566e-06, "loss": 0.2124, "step": 2956 }, { "epoch": 0.92, "learning_rate": 8.820164032806562e-06, "loss": 0.282, "step": 2957 }, { "epoch": 0.92, "learning_rate": 8.819763952790559e-06, "loss": 0.2954, "step": 2958 }, { "epoch": 0.92, "learning_rate": 8.819363872774556e-06, "loss": 0.2337, "step": 2959 }, { "epoch": 0.92, "learning_rate": 8.818963792758552e-06, "loss": 0.1941, "step": 2960 }, { "epoch": 0.92, "learning_rate": 8.818563712742549e-06, "loss": 0.2609, "step": 2961 }, { "epoch": 0.92, "learning_rate": 8.818163632726546e-06, "loss": 0.204, "step": 2962 }, { "epoch": 0.92, "learning_rate": 8.817763552710542e-06, "loss": 0.223, "step": 2963 }, { "epoch": 0.92, "learning_rate": 8.817363472694541e-06, "loss": 0.3096, "step": 2964 }, { "epoch": 0.92, "learning_rate": 8.816963392678538e-06, "loss": 0.164, "step": 2965 }, { "epoch": 0.92, "learning_rate": 8.816563312662533e-06, "loss": 0.2312, "step": 2966 }, { "epoch": 0.92, "learning_rate": 8.81616323264653e-06, "loss": 0.1834, "step": 2967 }, { "epoch": 0.92, "learning_rate": 8.815763152630526e-06, "loss": 0.1761, "step": 2968 }, { "epoch": 0.92, "learning_rate": 8.815363072614523e-06, "loss": 0.1979, "step": 2969 }, { "epoch": 0.92, "learning_rate": 8.814962992598521e-06, "loss": 0.3023, "step": 2970 }, { "epoch": 0.92, "learning_rate": 8.814562912582518e-06, "loss": 0.2656, "step": 2971 }, { "epoch": 0.92, "learning_rate": 8.814162832566514e-06, "loss": 0.2691, "step": 2972 }, { "epoch": 0.92, "learning_rate": 8.813762752550511e-06, "loss": 0.1332, "step": 2973 }, { "epoch": 0.92, "learning_rate": 8.813362672534508e-06, "loss": 0.3351, "step": 2974 }, { "epoch": 0.92, "learning_rate": 8.812962592518504e-06, "loss": 0.4034, "step": 2975 }, { "epoch": 0.92, "learning_rate": 8.812562512502501e-06, "loss": 0.2995, "step": 2976 }, { "epoch": 0.92, "learning_rate": 8.812162432486498e-06, "loss": 0.228, "step": 2977 }, { "epoch": 0.92, "learning_rate": 8.811762352470494e-06, "loss": 0.3139, "step": 2978 }, { "epoch": 0.92, "learning_rate": 8.811362272454491e-06, "loss": 0.1842, "step": 2979 }, { "epoch": 0.92, "learning_rate": 8.810962192438488e-06, "loss": 0.2103, "step": 2980 }, { "epoch": 0.93, "learning_rate": 8.810562112422485e-06, "loss": 0.4178, "step": 2981 }, { "epoch": 0.93, "learning_rate": 8.810162032406481e-06, "loss": 0.3238, "step": 2982 }, { "epoch": 0.93, "learning_rate": 8.809761952390478e-06, "loss": 0.2908, "step": 2983 }, { "epoch": 0.93, "learning_rate": 8.809361872374476e-06, "loss": 0.2694, "step": 2984 }, { "epoch": 0.93, "learning_rate": 8.808961792358473e-06, "loss": 0.285, "step": 2985 }, { "epoch": 0.93, "learning_rate": 8.80856171234247e-06, "loss": 0.2241, "step": 2986 }, { "epoch": 0.93, "learning_rate": 8.808161632326466e-06, "loss": 0.1584, "step": 2987 }, { "epoch": 0.93, "learning_rate": 8.807761552310463e-06, "loss": 0.1386, "step": 2988 }, { "epoch": 0.93, "learning_rate": 8.80736147229446e-06, "loss": 0.3377, "step": 2989 }, { "epoch": 0.93, "learning_rate": 8.806961392278456e-06, "loss": 0.3343, "step": 2990 }, { "epoch": 0.93, "learning_rate": 8.806561312262453e-06, "loss": 0.165, "step": 2991 }, { "epoch": 0.93, "learning_rate": 8.80616123224645e-06, "loss": 0.2675, "step": 2992 }, { "epoch": 0.93, "learning_rate": 8.805761152230446e-06, "loss": 0.2647, "step": 2993 }, { "epoch": 0.93, "learning_rate": 8.805361072214443e-06, "loss": 0.2123, "step": 2994 }, { "epoch": 0.93, "learning_rate": 8.80496099219844e-06, "loss": 0.2433, "step": 2995 }, { "epoch": 0.93, "learning_rate": 8.804560912182437e-06, "loss": 0.3286, "step": 2996 }, { "epoch": 0.93, "learning_rate": 8.804160832166433e-06, "loss": 0.1689, "step": 2997 }, { "epoch": 0.93, "learning_rate": 8.803760752150432e-06, "loss": 0.2292, "step": 2998 }, { "epoch": 0.93, "learning_rate": 8.803360672134428e-06, "loss": 0.2459, "step": 2999 }, { "epoch": 0.93, "learning_rate": 8.802960592118425e-06, "loss": 0.2194, "step": 3000 }, { "epoch": 0.93, "learning_rate": 8.802560512102422e-06, "loss": 0.1587, "step": 3001 }, { "epoch": 0.93, "learning_rate": 8.802160432086418e-06, "loss": 0.1776, "step": 3002 }, { "epoch": 0.93, "learning_rate": 8.801760352070415e-06, "loss": 0.4078, "step": 3003 }, { "epoch": 0.93, "learning_rate": 8.801360272054412e-06, "loss": 0.1398, "step": 3004 }, { "epoch": 0.93, "learning_rate": 8.800960192038408e-06, "loss": 0.2755, "step": 3005 }, { "epoch": 0.93, "learning_rate": 8.800560112022405e-06, "loss": 0.1924, "step": 3006 }, { "epoch": 0.93, "learning_rate": 8.800160032006402e-06, "loss": 0.3938, "step": 3007 }, { "epoch": 0.93, "learning_rate": 8.799759951990398e-06, "loss": 0.1225, "step": 3008 }, { "epoch": 0.93, "learning_rate": 8.799359871974395e-06, "loss": 0.2355, "step": 3009 }, { "epoch": 0.93, "learning_rate": 8.798959791958392e-06, "loss": 0.2696, "step": 3010 }, { "epoch": 0.93, "learning_rate": 8.798559711942389e-06, "loss": 0.3275, "step": 3011 }, { "epoch": 0.93, "learning_rate": 8.798159631926387e-06, "loss": 0.3877, "step": 3012 }, { "epoch": 0.94, "learning_rate": 8.797759551910384e-06, "loss": 0.282, "step": 3013 }, { "epoch": 0.94, "learning_rate": 8.79735947189438e-06, "loss": 0.3114, "step": 3014 }, { "epoch": 0.94, "learning_rate": 8.796959391878377e-06, "loss": 0.2474, "step": 3015 }, { "epoch": 0.94, "learning_rate": 8.796559311862372e-06, "loss": 0.1319, "step": 3016 }, { "epoch": 0.94, "learning_rate": 8.79615923184637e-06, "loss": 0.5265, "step": 3017 }, { "epoch": 0.94, "learning_rate": 8.795759151830367e-06, "loss": 0.201, "step": 3018 }, { "epoch": 0.94, "learning_rate": 8.795359071814364e-06, "loss": 0.2557, "step": 3019 }, { "epoch": 0.94, "learning_rate": 8.79495899179836e-06, "loss": 0.5792, "step": 3020 }, { "epoch": 0.94, "learning_rate": 8.794558911782357e-06, "loss": 0.2445, "step": 3021 }, { "epoch": 0.94, "learning_rate": 8.794158831766354e-06, "loss": 0.2633, "step": 3022 }, { "epoch": 0.94, "learning_rate": 8.79375875175035e-06, "loss": 0.228, "step": 3023 }, { "epoch": 0.94, "learning_rate": 8.793358671734347e-06, "loss": 0.1483, "step": 3024 }, { "epoch": 0.94, "learning_rate": 8.792958591718344e-06, "loss": 0.1414, "step": 3025 }, { "epoch": 0.94, "learning_rate": 8.792558511702342e-06, "loss": 0.2449, "step": 3026 }, { "epoch": 0.94, "learning_rate": 8.792158431686339e-06, "loss": 0.1826, "step": 3027 }, { "epoch": 0.94, "learning_rate": 8.791758351670336e-06, "loss": 0.2003, "step": 3028 }, { "epoch": 0.94, "learning_rate": 8.79135827165433e-06, "loss": 0.1942, "step": 3029 }, { "epoch": 0.94, "learning_rate": 8.790958191638327e-06, "loss": 0.3043, "step": 3030 }, { "epoch": 0.94, "learning_rate": 8.790558111622326e-06, "loss": 0.3753, "step": 3031 }, { "epoch": 0.94, "learning_rate": 8.790158031606322e-06, "loss": 0.2122, "step": 3032 }, { "epoch": 0.94, "learning_rate": 8.789757951590319e-06, "loss": 0.264, "step": 3033 }, { "epoch": 0.94, "learning_rate": 8.789357871574316e-06, "loss": 0.2794, "step": 3034 }, { "epoch": 0.94, "learning_rate": 8.788957791558312e-06, "loss": 0.2905, "step": 3035 }, { "epoch": 0.94, "learning_rate": 8.788557711542309e-06, "loss": 0.1738, "step": 3036 }, { "epoch": 0.94, "learning_rate": 8.788157631526306e-06, "loss": 0.3095, "step": 3037 }, { "epoch": 0.94, "learning_rate": 8.787757551510302e-06, "loss": 0.1808, "step": 3038 }, { "epoch": 0.94, "learning_rate": 8.787357471494299e-06, "loss": 0.3579, "step": 3039 }, { "epoch": 0.94, "learning_rate": 8.786957391478298e-06, "loss": 0.2787, "step": 3040 }, { "epoch": 0.94, "learning_rate": 8.786557311462292e-06, "loss": 0.1903, "step": 3041 }, { "epoch": 0.94, "learning_rate": 8.78615723144629e-06, "loss": 0.2554, "step": 3042 }, { "epoch": 0.94, "learning_rate": 8.785757151430286e-06, "loss": 0.1827, "step": 3043 }, { "epoch": 0.94, "learning_rate": 8.785357071414283e-06, "loss": 0.2271, "step": 3044 }, { "epoch": 0.95, "learning_rate": 8.784956991398281e-06, "loss": 0.322, "step": 3045 }, { "epoch": 0.95, "learning_rate": 8.784556911382278e-06, "loss": 0.1882, "step": 3046 }, { "epoch": 0.95, "learning_rate": 8.784156831366274e-06, "loss": 0.2229, "step": 3047 }, { "epoch": 0.95, "learning_rate": 8.783756751350271e-06, "loss": 0.1325, "step": 3048 }, { "epoch": 0.95, "learning_rate": 8.783356671334268e-06, "loss": 0.2539, "step": 3049 }, { "epoch": 0.95, "learning_rate": 8.782956591318264e-06, "loss": 0.2873, "step": 3050 }, { "epoch": 0.95, "learning_rate": 8.782556511302261e-06, "loss": 0.1472, "step": 3051 }, { "epoch": 0.95, "learning_rate": 8.782156431286258e-06, "loss": 0.2375, "step": 3052 }, { "epoch": 0.95, "learning_rate": 8.781756351270256e-06, "loss": 0.2139, "step": 3053 }, { "epoch": 0.95, "learning_rate": 8.781356271254251e-06, "loss": 0.2139, "step": 3054 }, { "epoch": 0.95, "learning_rate": 8.780956191238248e-06, "loss": 0.1264, "step": 3055 }, { "epoch": 0.95, "learning_rate": 8.780556111222244e-06, "loss": 0.2092, "step": 3056 }, { "epoch": 0.95, "learning_rate": 8.780156031206241e-06, "loss": 0.2157, "step": 3057 }, { "epoch": 0.95, "learning_rate": 8.779755951190238e-06, "loss": 0.2949, "step": 3058 }, { "epoch": 0.95, "learning_rate": 8.779355871174236e-06, "loss": 0.1466, "step": 3059 }, { "epoch": 0.95, "learning_rate": 8.778955791158233e-06, "loss": 0.2305, "step": 3060 }, { "epoch": 0.95, "learning_rate": 8.77855571114223e-06, "loss": 0.294, "step": 3061 }, { "epoch": 0.95, "learning_rate": 8.778155631126226e-06, "loss": 0.5309, "step": 3062 }, { "epoch": 0.95, "learning_rate": 8.777755551110223e-06, "loss": 0.2189, "step": 3063 }, { "epoch": 0.95, "learning_rate": 8.77735547109422e-06, "loss": 0.2486, "step": 3064 }, { "epoch": 0.95, "learning_rate": 8.776955391078216e-06, "loss": 0.3124, "step": 3065 }, { "epoch": 0.95, "learning_rate": 8.776555311062213e-06, "loss": 0.1904, "step": 3066 }, { "epoch": 0.95, "learning_rate": 8.77615523104621e-06, "loss": 0.4889, "step": 3067 }, { "epoch": 0.95, "learning_rate": 8.775755151030206e-06, "loss": 0.3491, "step": 3068 }, { "epoch": 0.95, "learning_rate": 8.775355071014203e-06, "loss": 0.3075, "step": 3069 }, { "epoch": 0.95, "learning_rate": 8.7749549909982e-06, "loss": 0.4136, "step": 3070 }, { "epoch": 0.95, "learning_rate": 8.774554910982196e-06, "loss": 0.117, "step": 3071 }, { "epoch": 0.95, "learning_rate": 8.774154830966193e-06, "loss": 0.2224, "step": 3072 }, { "epoch": 0.95, "learning_rate": 8.773754750950192e-06, "loss": 0.2067, "step": 3073 }, { "epoch": 0.95, "learning_rate": 8.773354670934188e-06, "loss": 0.1816, "step": 3074 }, { "epoch": 0.95, "learning_rate": 8.772954590918185e-06, "loss": 0.2459, "step": 3075 }, { "epoch": 0.95, "learning_rate": 8.772554510902182e-06, "loss": 0.2204, "step": 3076 }, { "epoch": 0.96, "learning_rate": 8.772154430886178e-06, "loss": 0.2721, "step": 3077 }, { "epoch": 0.96, "learning_rate": 8.771754350870175e-06, "loss": 0.1788, "step": 3078 }, { "epoch": 0.96, "learning_rate": 8.771354270854172e-06, "loss": 0.2882, "step": 3079 }, { "epoch": 0.96, "learning_rate": 8.770954190838168e-06, "loss": 0.1781, "step": 3080 }, { "epoch": 0.96, "learning_rate": 8.770554110822165e-06, "loss": 0.3593, "step": 3081 }, { "epoch": 0.96, "learning_rate": 8.770154030806162e-06, "loss": 0.2347, "step": 3082 }, { "epoch": 0.96, "learning_rate": 8.769753950790158e-06, "loss": 0.246, "step": 3083 }, { "epoch": 0.96, "learning_rate": 8.769353870774155e-06, "loss": 0.186, "step": 3084 }, { "epoch": 0.96, "learning_rate": 8.768953790758152e-06, "loss": 0.3362, "step": 3085 }, { "epoch": 0.96, "learning_rate": 8.768553710742148e-06, "loss": 0.2621, "step": 3086 }, { "epoch": 0.96, "learning_rate": 8.768153630726147e-06, "loss": 0.3886, "step": 3087 }, { "epoch": 0.96, "learning_rate": 8.767753550710144e-06, "loss": 0.3479, "step": 3088 }, { "epoch": 0.96, "learning_rate": 8.76735347069414e-06, "loss": 0.2144, "step": 3089 }, { "epoch": 0.96, "learning_rate": 8.766953390678137e-06, "loss": 0.2802, "step": 3090 }, { "epoch": 0.96, "learning_rate": 8.766553310662134e-06, "loss": 0.2828, "step": 3091 }, { "epoch": 0.96, "learning_rate": 8.76615323064613e-06, "loss": 0.2911, "step": 3092 }, { "epoch": 0.96, "learning_rate": 8.765753150630127e-06, "loss": 0.2717, "step": 3093 }, { "epoch": 0.96, "learning_rate": 8.765353070614124e-06, "loss": 0.2939, "step": 3094 }, { "epoch": 0.96, "learning_rate": 8.76495299059812e-06, "loss": 0.2175, "step": 3095 }, { "epoch": 0.96, "learning_rate": 8.764552910582117e-06, "loss": 0.2973, "step": 3096 }, { "epoch": 0.96, "learning_rate": 8.764152830566114e-06, "loss": 0.3981, "step": 3097 }, { "epoch": 0.96, "learning_rate": 8.76375275055011e-06, "loss": 0.2235, "step": 3098 }, { "epoch": 0.96, "learning_rate": 8.763352670534107e-06, "loss": 0.2146, "step": 3099 }, { "epoch": 0.96, "learning_rate": 8.762952590518104e-06, "loss": 0.213, "step": 3100 }, { "epoch": 0.96, "learning_rate": 8.762552510502102e-06, "loss": 0.2875, "step": 3101 }, { "epoch": 0.96, "learning_rate": 8.762152430486099e-06, "loss": 0.1243, "step": 3102 }, { "epoch": 0.96, "learning_rate": 8.761752350470096e-06, "loss": 0.2557, "step": 3103 }, { "epoch": 0.96, "learning_rate": 8.761352270454092e-06, "loss": 0.3246, "step": 3104 }, { "epoch": 0.96, "learning_rate": 8.760952190438087e-06, "loss": 0.3988, "step": 3105 }, { "epoch": 0.96, "learning_rate": 8.760552110422086e-06, "loss": 0.1752, "step": 3106 }, { "epoch": 0.96, "learning_rate": 8.760152030406082e-06, "loss": 0.4612, "step": 3107 }, { "epoch": 0.96, "learning_rate": 8.759751950390079e-06, "loss": 0.198, "step": 3108 }, { "epoch": 0.96, "learning_rate": 8.759351870374076e-06, "loss": 0.1921, "step": 3109 }, { "epoch": 0.97, "learning_rate": 8.758951790358072e-06, "loss": 0.456, "step": 3110 }, { "epoch": 0.97, "learning_rate": 8.758551710342069e-06, "loss": 0.1839, "step": 3111 }, { "epoch": 0.97, "learning_rate": 8.758151630326066e-06, "loss": 0.276, "step": 3112 }, { "epoch": 0.97, "learning_rate": 8.757751550310062e-06, "loss": 0.2925, "step": 3113 }, { "epoch": 0.97, "learning_rate": 8.757351470294059e-06, "loss": 0.2995, "step": 3114 }, { "epoch": 0.97, "learning_rate": 8.756951390278057e-06, "loss": 0.1858, "step": 3115 }, { "epoch": 0.97, "learning_rate": 8.756551310262054e-06, "loss": 0.1623, "step": 3116 }, { "epoch": 0.97, "learning_rate": 8.756151230246049e-06, "loss": 0.2021, "step": 3117 }, { "epoch": 0.97, "learning_rate": 8.755751150230046e-06, "loss": 0.2833, "step": 3118 }, { "epoch": 0.97, "learning_rate": 8.755351070214042e-06, "loss": 0.2051, "step": 3119 }, { "epoch": 0.97, "learning_rate": 8.754950990198041e-06, "loss": 0.1829, "step": 3120 }, { "epoch": 0.97, "learning_rate": 8.754550910182038e-06, "loss": 0.1485, "step": 3121 }, { "epoch": 0.97, "learning_rate": 8.754150830166034e-06, "loss": 0.1686, "step": 3122 }, { "epoch": 0.97, "learning_rate": 8.753750750150031e-06, "loss": 0.203, "step": 3123 }, { "epoch": 0.97, "learning_rate": 8.753350670134028e-06, "loss": 0.261, "step": 3124 }, { "epoch": 0.97, "learning_rate": 8.752950590118024e-06, "loss": 0.3254, "step": 3125 }, { "epoch": 0.97, "learning_rate": 8.752550510102021e-06, "loss": 0.1931, "step": 3126 }, { "epoch": 0.97, "learning_rate": 8.752150430086018e-06, "loss": 0.2726, "step": 3127 }, { "epoch": 0.97, "learning_rate": 8.751750350070014e-06, "loss": 0.2038, "step": 3128 }, { "epoch": 0.97, "learning_rate": 8.751350270054013e-06, "loss": 0.1363, "step": 3129 }, { "epoch": 0.97, "learning_rate": 8.750950190038008e-06, "loss": 0.2364, "step": 3130 }, { "epoch": 0.97, "learning_rate": 8.750550110022004e-06, "loss": 0.2457, "step": 3131 }, { "epoch": 0.97, "learning_rate": 8.750150030006001e-06, "loss": 0.3498, "step": 3132 }, { "epoch": 0.97, "learning_rate": 8.749749949989998e-06, "loss": 0.2168, "step": 3133 }, { "epoch": 0.97, "learning_rate": 8.749349869973996e-06, "loss": 0.2183, "step": 3134 }, { "epoch": 0.97, "learning_rate": 8.748949789957993e-06, "loss": 0.2784, "step": 3135 }, { "epoch": 0.97, "learning_rate": 8.74854970994199e-06, "loss": 0.22, "step": 3136 }, { "epoch": 0.97, "learning_rate": 8.748149629925986e-06, "loss": 0.2216, "step": 3137 }, { "epoch": 0.97, "learning_rate": 8.747749549909983e-06, "loss": 0.212, "step": 3138 }, { "epoch": 0.97, "learning_rate": 8.74734946989398e-06, "loss": 0.2099, "step": 3139 }, { "epoch": 0.97, "learning_rate": 8.746949389877976e-06, "loss": 0.29, "step": 3140 }, { "epoch": 0.97, "learning_rate": 8.746549309861973e-06, "loss": 0.2584, "step": 3141 }, { "epoch": 0.98, "learning_rate": 8.74614922984597e-06, "loss": 0.1769, "step": 3142 }, { "epoch": 0.98, "learning_rate": 8.745749149829966e-06, "loss": 0.1742, "step": 3143 }, { "epoch": 0.98, "learning_rate": 8.745349069813963e-06, "loss": 0.1136, "step": 3144 }, { "epoch": 0.98, "learning_rate": 8.74494898979796e-06, "loss": 0.1918, "step": 3145 }, { "epoch": 0.98, "learning_rate": 8.744548909781956e-06, "loss": 0.1946, "step": 3146 }, { "epoch": 0.98, "learning_rate": 8.744148829765953e-06, "loss": 0.4583, "step": 3147 }, { "epoch": 0.98, "learning_rate": 8.743748749749951e-06, "loss": 0.1683, "step": 3148 }, { "epoch": 0.98, "learning_rate": 8.743348669733948e-06, "loss": 0.2649, "step": 3149 }, { "epoch": 0.98, "learning_rate": 8.742948589717945e-06, "loss": 0.2586, "step": 3150 }, { "epoch": 0.98, "learning_rate": 8.742548509701942e-06, "loss": 0.1956, "step": 3151 }, { "epoch": 0.98, "learning_rate": 8.742148429685938e-06, "loss": 0.3042, "step": 3152 }, { "epoch": 0.98, "learning_rate": 8.741748349669935e-06, "loss": 0.1587, "step": 3153 }, { "epoch": 0.98, "learning_rate": 8.741348269653932e-06, "loss": 0.1951, "step": 3154 }, { "epoch": 0.98, "learning_rate": 8.740948189637928e-06, "loss": 0.2161, "step": 3155 }, { "epoch": 0.98, "learning_rate": 8.740548109621925e-06, "loss": 0.2286, "step": 3156 }, { "epoch": 0.98, "learning_rate": 8.740148029605922e-06, "loss": 0.1549, "step": 3157 }, { "epoch": 0.98, "learning_rate": 8.739747949589918e-06, "loss": 0.2951, "step": 3158 }, { "epoch": 0.98, "learning_rate": 8.739347869573915e-06, "loss": 0.2218, "step": 3159 }, { "epoch": 0.98, "learning_rate": 8.738947789557912e-06, "loss": 0.1401, "step": 3160 }, { "epoch": 0.98, "learning_rate": 8.738547709541908e-06, "loss": 0.4156, "step": 3161 }, { "epoch": 0.98, "learning_rate": 8.738147629525907e-06, "loss": 0.3701, "step": 3162 }, { "epoch": 0.98, "learning_rate": 8.737747549509903e-06, "loss": 0.2875, "step": 3163 }, { "epoch": 0.98, "learning_rate": 8.7373474694939e-06, "loss": 0.1854, "step": 3164 }, { "epoch": 0.98, "learning_rate": 8.736947389477897e-06, "loss": 0.2321, "step": 3165 }, { "epoch": 0.98, "learning_rate": 8.736547309461894e-06, "loss": 0.2091, "step": 3166 }, { "epoch": 0.98, "learning_rate": 8.73614722944589e-06, "loss": 0.3833, "step": 3167 }, { "epoch": 0.98, "learning_rate": 8.735747149429887e-06, "loss": 0.1514, "step": 3168 }, { "epoch": 0.98, "learning_rate": 8.735347069413884e-06, "loss": 0.4566, "step": 3169 }, { "epoch": 0.98, "learning_rate": 8.73494698939788e-06, "loss": 0.1275, "step": 3170 }, { "epoch": 0.98, "learning_rate": 8.734546909381877e-06, "loss": 0.1991, "step": 3171 }, { "epoch": 0.98, "learning_rate": 8.734146829365874e-06, "loss": 0.2041, "step": 3172 }, { "epoch": 0.98, "learning_rate": 8.73374674934987e-06, "loss": 0.1893, "step": 3173 }, { "epoch": 0.99, "learning_rate": 8.733346669333867e-06, "loss": 0.1457, "step": 3174 }, { "epoch": 0.99, "learning_rate": 8.732946589317864e-06, "loss": 0.3085, "step": 3175 }, { "epoch": 0.99, "learning_rate": 8.732546509301862e-06, "loss": 0.2002, "step": 3176 }, { "epoch": 0.99, "learning_rate": 8.732146429285859e-06, "loss": 0.2195, "step": 3177 }, { "epoch": 0.99, "learning_rate": 8.731746349269855e-06, "loss": 0.2963, "step": 3178 }, { "epoch": 0.99, "learning_rate": 8.731346269253852e-06, "loss": 0.1813, "step": 3179 }, { "epoch": 0.99, "learning_rate": 8.730946189237847e-06, "loss": 0.1875, "step": 3180 }, { "epoch": 0.99, "learning_rate": 8.730546109221844e-06, "loss": 0.2851, "step": 3181 }, { "epoch": 0.99, "learning_rate": 8.730146029205842e-06, "loss": 0.4599, "step": 3182 }, { "epoch": 0.99, "learning_rate": 8.729745949189839e-06, "loss": 0.2975, "step": 3183 }, { "epoch": 0.99, "learning_rate": 8.729345869173836e-06, "loss": 0.1789, "step": 3184 }, { "epoch": 0.99, "learning_rate": 8.728945789157832e-06, "loss": 0.1138, "step": 3185 }, { "epoch": 0.99, "learning_rate": 8.728545709141829e-06, "loss": 0.1399, "step": 3186 }, { "epoch": 0.99, "learning_rate": 8.728145629125826e-06, "loss": 0.1602, "step": 3187 }, { "epoch": 0.99, "learning_rate": 8.727745549109822e-06, "loss": 0.3891, "step": 3188 }, { "epoch": 0.99, "learning_rate": 8.727345469093819e-06, "loss": 0.3135, "step": 3189 }, { "epoch": 0.99, "learning_rate": 8.726945389077817e-06, "loss": 0.3129, "step": 3190 }, { "epoch": 0.99, "learning_rate": 8.726545309061814e-06, "loss": 0.191, "step": 3191 }, { "epoch": 0.99, "learning_rate": 8.72614522904581e-06, "loss": 0.1978, "step": 3192 }, { "epoch": 0.99, "learning_rate": 8.725745149029806e-06, "loss": 0.2209, "step": 3193 }, { "epoch": 0.99, "learning_rate": 8.725345069013802e-06, "loss": 0.4304, "step": 3194 }, { "epoch": 0.99, "learning_rate": 8.724944988997799e-06, "loss": 0.2078, "step": 3195 }, { "epoch": 0.99, "learning_rate": 8.724544908981798e-06, "loss": 0.3345, "step": 3196 }, { "epoch": 0.99, "learning_rate": 8.724144828965794e-06, "loss": 0.1868, "step": 3197 }, { "epoch": 0.99, "learning_rate": 8.723744748949791e-06, "loss": 0.3092, "step": 3198 }, { "epoch": 0.99, "learning_rate": 8.723344668933788e-06, "loss": 0.1601, "step": 3199 }, { "epoch": 0.99, "learning_rate": 8.722944588917784e-06, "loss": 0.371, "step": 3200 }, { "epoch": 0.99, "learning_rate": 8.722544508901781e-06, "loss": 0.3004, "step": 3201 }, { "epoch": 0.99, "learning_rate": 8.722144428885778e-06, "loss": 0.2874, "step": 3202 }, { "epoch": 0.99, "learning_rate": 8.721744348869774e-06, "loss": 0.3302, "step": 3203 }, { "epoch": 0.99, "learning_rate": 8.721344268853773e-06, "loss": 0.2971, "step": 3204 }, { "epoch": 0.99, "learning_rate": 8.720944188837768e-06, "loss": 0.3019, "step": 3205 }, { "epoch": 1.0, "learning_rate": 8.720544108821764e-06, "loss": 0.2954, "step": 3206 }, { "epoch": 1.0, "learning_rate": 8.720144028805761e-06, "loss": 0.2214, "step": 3207 }, { "epoch": 1.0, "learning_rate": 8.719743948789758e-06, "loss": 0.1656, "step": 3208 }, { "epoch": 1.0, "learning_rate": 8.719343868773754e-06, "loss": 0.1975, "step": 3209 }, { "epoch": 1.0, "learning_rate": 8.718943788757753e-06, "loss": 0.1764, "step": 3210 }, { "epoch": 1.0, "learning_rate": 8.71854370874175e-06, "loss": 0.2486, "step": 3211 }, { "epoch": 1.0, "learning_rate": 8.718143628725746e-06, "loss": 0.3554, "step": 3212 }, { "epoch": 1.0, "learning_rate": 8.717743548709743e-06, "loss": 0.2227, "step": 3213 }, { "epoch": 1.0, "learning_rate": 8.71734346869374e-06, "loss": 0.2242, "step": 3214 }, { "epoch": 1.0, "learning_rate": 8.716943388677736e-06, "loss": 0.2421, "step": 3215 }, { "epoch": 1.0, "learning_rate": 8.716543308661733e-06, "loss": 0.1381, "step": 3216 }, { "epoch": 1.0, "learning_rate": 8.71614322864573e-06, "loss": 0.3306, "step": 3217 }, { "epoch": 1.0, "learning_rate": 8.715743148629726e-06, "loss": 0.1816, "step": 3218 }, { "epoch": 1.0, "learning_rate": 8.715343068613723e-06, "loss": 0.2269, "step": 3219 }, { "epoch": 1.0, "learning_rate": 8.71494298859772e-06, "loss": 0.4335, "step": 3220 }, { "epoch": 1.0, "learning_rate": 8.714542908581716e-06, "loss": 0.1668, "step": 3221 }, { "epoch": 1.0, "learning_rate": 8.714142828565713e-06, "loss": 0.1296, "step": 3222 }, { "epoch": 1.0, "learning_rate": 8.713742748549711e-06, "loss": 0.18, "step": 3223 }, { "epoch": 1.0, "learning_rate": 8.713342668533708e-06, "loss": 0.2465, "step": 3224 }, { "epoch": 1.0, "learning_rate": 8.712942588517705e-06, "loss": 0.1434, "step": 3225 }, { "epoch": 1.0, "learning_rate": 8.712542508501701e-06, "loss": 0.1268, "step": 3226 }, { "epoch": 1.0, "learning_rate": 8.712142428485698e-06, "loss": 0.1062, "step": 3227 }, { "epoch": 1.0, "learning_rate": 8.711742348469695e-06, "loss": 0.134, "step": 3228 }, { "epoch": 1.0, "learning_rate": 8.711342268453692e-06, "loss": 0.1631, "step": 3229 }, { "epoch": 1.0, "learning_rate": 8.710942188437688e-06, "loss": 0.2786, "step": 3230 }, { "epoch": 1.0, "learning_rate": 8.710542108421685e-06, "loss": 0.139, "step": 3231 }, { "epoch": 1.0, "learning_rate": 8.710142028405682e-06, "loss": 0.3208, "step": 3232 }, { "epoch": 1.0, "learning_rate": 8.709741948389678e-06, "loss": 0.165, "step": 3233 }, { "epoch": 1.0, "learning_rate": 8.709341868373675e-06, "loss": 0.1482, "step": 3234 }, { "epoch": 1.0, "learning_rate": 8.708941788357672e-06, "loss": 0.1194, "step": 3235 }, { "epoch": 1.0, "learning_rate": 8.708541708341668e-06, "loss": 0.1007, "step": 3236 }, { "epoch": 1.0, "learning_rate": 8.708141628325667e-06, "loss": 0.1605, "step": 3237 }, { "epoch": 1.01, "learning_rate": 8.707741548309663e-06, "loss": 0.3381, "step": 3238 }, { "epoch": 1.01, "learning_rate": 8.70734146829366e-06, "loss": 0.1209, "step": 3239 }, { "epoch": 1.01, "learning_rate": 8.706941388277657e-06, "loss": 0.1666, "step": 3240 }, { "epoch": 1.01, "learning_rate": 8.706541308261653e-06, "loss": 0.2026, "step": 3241 }, { "epoch": 1.01, "learning_rate": 8.70614122824565e-06, "loss": 0.1311, "step": 3242 }, { "epoch": 1.01, "learning_rate": 8.705741148229647e-06, "loss": 0.1202, "step": 3243 }, { "epoch": 1.01, "learning_rate": 8.705341068213644e-06, "loss": 0.1439, "step": 3244 }, { "epoch": 1.01, "learning_rate": 8.70494098819764e-06, "loss": 0.15, "step": 3245 }, { "epoch": 1.01, "learning_rate": 8.704540908181637e-06, "loss": 0.1673, "step": 3246 }, { "epoch": 1.01, "learning_rate": 8.704140828165634e-06, "loss": 0.2118, "step": 3247 }, { "epoch": 1.01, "learning_rate": 8.70374074814963e-06, "loss": 0.1235, "step": 3248 }, { "epoch": 1.01, "learning_rate": 8.703340668133627e-06, "loss": 0.2793, "step": 3249 }, { "epoch": 1.01, "learning_rate": 8.702940588117624e-06, "loss": 0.1934, "step": 3250 }, { "epoch": 1.01, "learning_rate": 8.702540508101622e-06, "loss": 0.1981, "step": 3251 }, { "epoch": 1.01, "learning_rate": 8.702140428085619e-06, "loss": 0.1397, "step": 3252 }, { "epoch": 1.01, "learning_rate": 8.701740348069615e-06, "loss": 0.1673, "step": 3253 }, { "epoch": 1.01, "learning_rate": 8.701340268053612e-06, "loss": 0.1137, "step": 3254 }, { "epoch": 1.01, "learning_rate": 8.700940188037609e-06, "loss": 0.0613, "step": 3255 }, { "epoch": 1.01, "learning_rate": 8.700540108021604e-06, "loss": 0.1336, "step": 3256 }, { "epoch": 1.01, "learning_rate": 8.700140028005602e-06, "loss": 0.1653, "step": 3257 }, { "epoch": 1.01, "learning_rate": 8.699739947989599e-06, "loss": 0.1613, "step": 3258 }, { "epoch": 1.01, "learning_rate": 8.699339867973596e-06, "loss": 0.1724, "step": 3259 }, { "epoch": 1.01, "learning_rate": 8.698939787957592e-06, "loss": 0.1917, "step": 3260 }, { "epoch": 1.01, "learning_rate": 8.698539707941589e-06, "loss": 0.1274, "step": 3261 }, { "epoch": 1.01, "learning_rate": 8.698139627925586e-06, "loss": 0.1617, "step": 3262 }, { "epoch": 1.01, "learning_rate": 8.697739547909582e-06, "loss": 0.2189, "step": 3263 }, { "epoch": 1.01, "learning_rate": 8.697339467893579e-06, "loss": 0.1112, "step": 3264 }, { "epoch": 1.01, "learning_rate": 8.696939387877577e-06, "loss": 0.1828, "step": 3265 }, { "epoch": 1.01, "learning_rate": 8.696539307861574e-06, "loss": 0.3767, "step": 3266 }, { "epoch": 1.01, "learning_rate": 8.69613922784557e-06, "loss": 0.1522, "step": 3267 }, { "epoch": 1.01, "learning_rate": 8.695739147829566e-06, "loss": 0.1883, "step": 3268 }, { "epoch": 1.01, "learning_rate": 8.695339067813562e-06, "loss": 0.1622, "step": 3269 }, { "epoch": 1.01, "learning_rate": 8.694938987797559e-06, "loss": 0.2092, "step": 3270 }, { "epoch": 1.02, "learning_rate": 8.694538907781557e-06, "loss": 0.1249, "step": 3271 }, { "epoch": 1.02, "learning_rate": 8.694138827765554e-06, "loss": 0.369, "step": 3272 }, { "epoch": 1.02, "learning_rate": 8.69373874774955e-06, "loss": 0.1118, "step": 3273 }, { "epoch": 1.02, "learning_rate": 8.693338667733548e-06, "loss": 0.145, "step": 3274 }, { "epoch": 1.02, "learning_rate": 8.692938587717544e-06, "loss": 0.1286, "step": 3275 }, { "epoch": 1.02, "learning_rate": 8.692538507701541e-06, "loss": 0.074, "step": 3276 }, { "epoch": 1.02, "learning_rate": 8.692138427685538e-06, "loss": 0.1773, "step": 3277 }, { "epoch": 1.02, "learning_rate": 8.691738347669534e-06, "loss": 0.0966, "step": 3278 }, { "epoch": 1.02, "learning_rate": 8.691338267653533e-06, "loss": 0.1415, "step": 3279 }, { "epoch": 1.02, "learning_rate": 8.69093818763753e-06, "loss": 0.1767, "step": 3280 }, { "epoch": 1.02, "learning_rate": 8.690538107621524e-06, "loss": 0.1469, "step": 3281 }, { "epoch": 1.02, "learning_rate": 8.690138027605521e-06, "loss": 0.158, "step": 3282 }, { "epoch": 1.02, "learning_rate": 8.689737947589518e-06, "loss": 0.176, "step": 3283 }, { "epoch": 1.02, "learning_rate": 8.689337867573514e-06, "loss": 0.1227, "step": 3284 }, { "epoch": 1.02, "learning_rate": 8.688937787557513e-06, "loss": 0.1195, "step": 3285 }, { "epoch": 1.02, "learning_rate": 8.68853770754151e-06, "loss": 0.2742, "step": 3286 }, { "epoch": 1.02, "learning_rate": 8.688137627525506e-06, "loss": 0.1884, "step": 3287 }, { "epoch": 1.02, "learning_rate": 8.687737547509503e-06, "loss": 0.3628, "step": 3288 }, { "epoch": 1.02, "learning_rate": 8.6873374674935e-06, "loss": 0.1594, "step": 3289 }, { "epoch": 1.02, "learning_rate": 8.686937387477496e-06, "loss": 0.1314, "step": 3290 }, { "epoch": 1.02, "learning_rate": 8.686537307461493e-06, "loss": 0.1592, "step": 3291 }, { "epoch": 1.02, "learning_rate": 8.68613722744549e-06, "loss": 0.0904, "step": 3292 }, { "epoch": 1.02, "learning_rate": 8.685737147429488e-06, "loss": 0.1094, "step": 3293 }, { "epoch": 1.02, "learning_rate": 8.685337067413483e-06, "loss": 0.1124, "step": 3294 }, { "epoch": 1.02, "learning_rate": 8.68493698739748e-06, "loss": 0.3009, "step": 3295 }, { "epoch": 1.02, "learning_rate": 8.684536907381476e-06, "loss": 0.157, "step": 3296 }, { "epoch": 1.02, "learning_rate": 8.684136827365473e-06, "loss": 0.1119, "step": 3297 }, { "epoch": 1.02, "learning_rate": 8.68373674734947e-06, "loss": 0.0956, "step": 3298 }, { "epoch": 1.02, "learning_rate": 8.683336667333468e-06, "loss": 0.1873, "step": 3299 }, { "epoch": 1.02, "learning_rate": 8.682936587317465e-06, "loss": 0.134, "step": 3300 }, { "epoch": 1.02, "learning_rate": 8.682536507301461e-06, "loss": 0.1815, "step": 3301 }, { "epoch": 1.02, "learning_rate": 8.682136427285458e-06, "loss": 0.2628, "step": 3302 }, { "epoch": 1.03, "learning_rate": 8.681736347269455e-06, "loss": 0.0938, "step": 3303 }, { "epoch": 1.03, "learning_rate": 8.681336267253451e-06, "loss": 0.1076, "step": 3304 }, { "epoch": 1.03, "learning_rate": 8.680936187237448e-06, "loss": 0.2398, "step": 3305 }, { "epoch": 1.03, "learning_rate": 8.680536107221445e-06, "loss": 0.1196, "step": 3306 }, { "epoch": 1.03, "learning_rate": 8.680136027205442e-06, "loss": 0.2009, "step": 3307 }, { "epoch": 1.03, "learning_rate": 8.679735947189438e-06, "loss": 0.1521, "step": 3308 }, { "epoch": 1.03, "learning_rate": 8.679335867173435e-06, "loss": 0.2214, "step": 3309 }, { "epoch": 1.03, "learning_rate": 8.678935787157432e-06, "loss": 0.0765, "step": 3310 }, { "epoch": 1.03, "learning_rate": 8.678535707141428e-06, "loss": 0.1263, "step": 3311 }, { "epoch": 1.03, "learning_rate": 8.678135627125425e-06, "loss": 0.1714, "step": 3312 }, { "epoch": 1.03, "learning_rate": 8.677735547109423e-06, "loss": 0.2162, "step": 3313 }, { "epoch": 1.03, "learning_rate": 8.67733546709342e-06, "loss": 0.2321, "step": 3314 }, { "epoch": 1.03, "learning_rate": 8.676935387077417e-06, "loss": 0.1684, "step": 3315 }, { "epoch": 1.03, "learning_rate": 8.676535307061413e-06, "loss": 0.285, "step": 3316 }, { "epoch": 1.03, "learning_rate": 8.67613522704541e-06, "loss": 0.1644, "step": 3317 }, { "epoch": 1.03, "learning_rate": 8.675735147029407e-06, "loss": 0.1392, "step": 3318 }, { "epoch": 1.03, "learning_rate": 8.675335067013403e-06, "loss": 0.1312, "step": 3319 }, { "epoch": 1.03, "learning_rate": 8.6749349869974e-06, "loss": 0.2034, "step": 3320 }, { "epoch": 1.03, "learning_rate": 8.674534906981397e-06, "loss": 0.2255, "step": 3321 }, { "epoch": 1.03, "learning_rate": 8.674134826965394e-06, "loss": 0.1905, "step": 3322 }, { "epoch": 1.03, "learning_rate": 8.67373474694939e-06, "loss": 0.208, "step": 3323 }, { "epoch": 1.03, "learning_rate": 8.673334666933387e-06, "loss": 0.1463, "step": 3324 }, { "epoch": 1.03, "learning_rate": 8.672934586917384e-06, "loss": 0.1384, "step": 3325 }, { "epoch": 1.03, "learning_rate": 8.67253450690138e-06, "loss": 0.1735, "step": 3326 }, { "epoch": 1.03, "learning_rate": 8.672134426885379e-06, "loss": 0.1837, "step": 3327 }, { "epoch": 1.03, "learning_rate": 8.671734346869375e-06, "loss": 0.2377, "step": 3328 }, { "epoch": 1.03, "learning_rate": 8.671334266853372e-06, "loss": 0.2128, "step": 3329 }, { "epoch": 1.03, "learning_rate": 8.670934186837369e-06, "loss": 0.134, "step": 3330 }, { "epoch": 1.03, "learning_rate": 8.670534106821364e-06, "loss": 0.1681, "step": 3331 }, { "epoch": 1.03, "learning_rate": 8.670134026805362e-06, "loss": 0.2334, "step": 3332 }, { "epoch": 1.03, "learning_rate": 8.669733946789359e-06, "loss": 0.3002, "step": 3333 }, { "epoch": 1.03, "learning_rate": 8.669333866773355e-06, "loss": 0.157, "step": 3334 }, { "epoch": 1.04, "learning_rate": 8.668933786757352e-06, "loss": 0.1846, "step": 3335 }, { "epoch": 1.04, "learning_rate": 8.668533706741349e-06, "loss": 0.2216, "step": 3336 }, { "epoch": 1.04, "learning_rate": 8.668133626725346e-06, "loss": 0.2056, "step": 3337 }, { "epoch": 1.04, "learning_rate": 8.667733546709342e-06, "loss": 0.1237, "step": 3338 }, { "epoch": 1.04, "learning_rate": 8.667333466693339e-06, "loss": 0.1482, "step": 3339 }, { "epoch": 1.04, "learning_rate": 8.666933386677336e-06, "loss": 0.2183, "step": 3340 }, { "epoch": 1.04, "learning_rate": 8.666533306661334e-06, "loss": 0.1843, "step": 3341 }, { "epoch": 1.04, "learning_rate": 8.66613322664533e-06, "loss": 0.2072, "step": 3342 }, { "epoch": 1.04, "learning_rate": 8.665733146629327e-06, "loss": 0.131, "step": 3343 }, { "epoch": 1.04, "learning_rate": 8.665333066613322e-06, "loss": 0.0877, "step": 3344 }, { "epoch": 1.04, "learning_rate": 8.664932986597319e-06, "loss": 0.104, "step": 3345 }, { "epoch": 1.04, "learning_rate": 8.664532906581317e-06, "loss": 0.1272, "step": 3346 }, { "epoch": 1.04, "learning_rate": 8.664132826565314e-06, "loss": 0.2476, "step": 3347 }, { "epoch": 1.04, "learning_rate": 8.66373274654931e-06, "loss": 0.2257, "step": 3348 }, { "epoch": 1.04, "learning_rate": 8.663332666533307e-06, "loss": 0.1255, "step": 3349 }, { "epoch": 1.04, "learning_rate": 8.662932586517304e-06, "loss": 0.1163, "step": 3350 }, { "epoch": 1.04, "learning_rate": 8.6625325065013e-06, "loss": 0.1937, "step": 3351 }, { "epoch": 1.04, "learning_rate": 8.662132426485298e-06, "loss": 0.184, "step": 3352 }, { "epoch": 1.04, "learning_rate": 8.661732346469294e-06, "loss": 0.1641, "step": 3353 }, { "epoch": 1.04, "learning_rate": 8.661332266453293e-06, "loss": 0.2236, "step": 3354 }, { "epoch": 1.04, "learning_rate": 8.66093218643729e-06, "loss": 0.1164, "step": 3355 }, { "epoch": 1.04, "learning_rate": 8.660532106421286e-06, "loss": 0.1408, "step": 3356 }, { "epoch": 1.04, "learning_rate": 8.660132026405281e-06, "loss": 0.1751, "step": 3357 }, { "epoch": 1.04, "learning_rate": 8.659731946389278e-06, "loss": 0.1092, "step": 3358 }, { "epoch": 1.04, "learning_rate": 8.659331866373274e-06, "loss": 0.2133, "step": 3359 }, { "epoch": 1.04, "learning_rate": 8.658931786357273e-06, "loss": 0.1452, "step": 3360 }, { "epoch": 1.04, "learning_rate": 8.65853170634127e-06, "loss": 0.3341, "step": 3361 }, { "epoch": 1.04, "learning_rate": 8.658131626325266e-06, "loss": 0.1753, "step": 3362 }, { "epoch": 1.04, "learning_rate": 8.657731546309263e-06, "loss": 0.2625, "step": 3363 }, { "epoch": 1.04, "learning_rate": 8.65733146629326e-06, "loss": 0.1804, "step": 3364 }, { "epoch": 1.04, "learning_rate": 8.656931386277256e-06, "loss": 0.1197, "step": 3365 }, { "epoch": 1.04, "learning_rate": 8.656531306261253e-06, "loss": 0.1568, "step": 3366 }, { "epoch": 1.05, "learning_rate": 8.65613122624525e-06, "loss": 0.2039, "step": 3367 }, { "epoch": 1.05, "learning_rate": 8.655731146229248e-06, "loss": 0.2546, "step": 3368 }, { "epoch": 1.05, "learning_rate": 8.655331066213243e-06, "loss": 0.2308, "step": 3369 }, { "epoch": 1.05, "learning_rate": 8.65493098619724e-06, "loss": 0.2908, "step": 3370 }, { "epoch": 1.05, "learning_rate": 8.654530906181236e-06, "loss": 0.2261, "step": 3371 }, { "epoch": 1.05, "learning_rate": 8.654130826165233e-06, "loss": 0.17, "step": 3372 }, { "epoch": 1.05, "learning_rate": 8.65373074614923e-06, "loss": 0.1013, "step": 3373 }, { "epoch": 1.05, "learning_rate": 8.653330666133228e-06, "loss": 0.1633, "step": 3374 }, { "epoch": 1.05, "learning_rate": 8.652930586117225e-06, "loss": 0.1498, "step": 3375 }, { "epoch": 1.05, "learning_rate": 8.652530506101221e-06, "loss": 0.2506, "step": 3376 }, { "epoch": 1.05, "learning_rate": 8.652130426085218e-06, "loss": 0.1103, "step": 3377 }, { "epoch": 1.05, "learning_rate": 8.651730346069215e-06, "loss": 0.1472, "step": 3378 }, { "epoch": 1.05, "learning_rate": 8.651330266053211e-06, "loss": 0.2295, "step": 3379 }, { "epoch": 1.05, "learning_rate": 8.650930186037208e-06, "loss": 0.1648, "step": 3380 }, { "epoch": 1.05, "learning_rate": 8.650530106021205e-06, "loss": 0.1898, "step": 3381 }, { "epoch": 1.05, "learning_rate": 8.650130026005201e-06, "loss": 0.1214, "step": 3382 }, { "epoch": 1.05, "learning_rate": 8.649729945989198e-06, "loss": 0.1687, "step": 3383 }, { "epoch": 1.05, "learning_rate": 8.649329865973195e-06, "loss": 0.1828, "step": 3384 }, { "epoch": 1.05, "learning_rate": 8.648929785957192e-06, "loss": 0.3385, "step": 3385 }, { "epoch": 1.05, "learning_rate": 8.648529705941188e-06, "loss": 0.1921, "step": 3386 }, { "epoch": 1.05, "learning_rate": 8.648129625925185e-06, "loss": 0.1903, "step": 3387 }, { "epoch": 1.05, "learning_rate": 8.647729545909183e-06, "loss": 0.3171, "step": 3388 }, { "epoch": 1.05, "learning_rate": 8.64732946589318e-06, "loss": 0.2232, "step": 3389 }, { "epoch": 1.05, "learning_rate": 8.646929385877177e-06, "loss": 0.2982, "step": 3390 }, { "epoch": 1.05, "learning_rate": 8.646529305861173e-06, "loss": 0.147, "step": 3391 }, { "epoch": 1.05, "learning_rate": 8.64612922584517e-06, "loss": 0.1541, "step": 3392 }, { "epoch": 1.05, "learning_rate": 8.645729145829167e-06, "loss": 0.1834, "step": 3393 }, { "epoch": 1.05, "learning_rate": 8.645329065813163e-06, "loss": 0.1004, "step": 3394 }, { "epoch": 1.05, "learning_rate": 8.64492898579716e-06, "loss": 0.1882, "step": 3395 }, { "epoch": 1.05, "learning_rate": 8.644528905781157e-06, "loss": 0.1311, "step": 3396 }, { "epoch": 1.05, "learning_rate": 8.644128825765153e-06, "loss": 0.1378, "step": 3397 }, { "epoch": 1.05, "learning_rate": 8.64372874574915e-06, "loss": 0.2101, "step": 3398 }, { "epoch": 1.05, "learning_rate": 8.643328665733147e-06, "loss": 0.1225, "step": 3399 }, { "epoch": 1.06, "learning_rate": 8.642928585717144e-06, "loss": 0.0743, "step": 3400 }, { "epoch": 1.06, "learning_rate": 8.64252850570114e-06, "loss": 0.1547, "step": 3401 }, { "epoch": 1.06, "learning_rate": 8.642128425685139e-06, "loss": 0.1332, "step": 3402 }, { "epoch": 1.06, "learning_rate": 8.641728345669135e-06, "loss": 0.1596, "step": 3403 }, { "epoch": 1.06, "learning_rate": 8.641328265653132e-06, "loss": 0.2137, "step": 3404 }, { "epoch": 1.06, "learning_rate": 8.640928185637129e-06, "loss": 0.3788, "step": 3405 }, { "epoch": 1.06, "learning_rate": 8.640528105621125e-06, "loss": 0.3002, "step": 3406 }, { "epoch": 1.06, "learning_rate": 8.640128025605122e-06, "loss": 0.3697, "step": 3407 }, { "epoch": 1.06, "learning_rate": 8.639727945589119e-06, "loss": 0.1567, "step": 3408 }, { "epoch": 1.06, "learning_rate": 8.639327865573115e-06, "loss": 0.1703, "step": 3409 }, { "epoch": 1.06, "learning_rate": 8.638927785557112e-06, "loss": 0.1555, "step": 3410 }, { "epoch": 1.06, "learning_rate": 8.638527705541109e-06, "loss": 0.2186, "step": 3411 }, { "epoch": 1.06, "learning_rate": 8.638127625525105e-06, "loss": 0.2017, "step": 3412 }, { "epoch": 1.06, "learning_rate": 8.637727545509102e-06, "loss": 0.1814, "step": 3413 }, { "epoch": 1.06, "learning_rate": 8.637327465493099e-06, "loss": 0.088, "step": 3414 }, { "epoch": 1.06, "learning_rate": 8.636927385477096e-06, "loss": 0.1169, "step": 3415 }, { "epoch": 1.06, "learning_rate": 8.636527305461094e-06, "loss": 0.0963, "step": 3416 }, { "epoch": 1.06, "learning_rate": 8.63612722544509e-06, "loss": 0.2396, "step": 3417 }, { "epoch": 1.06, "learning_rate": 8.635727145429087e-06, "loss": 0.2062, "step": 3418 }, { "epoch": 1.06, "learning_rate": 8.635327065413084e-06, "loss": 0.1245, "step": 3419 }, { "epoch": 1.06, "learning_rate": 8.634926985397079e-06, "loss": 0.2347, "step": 3420 }, { "epoch": 1.06, "learning_rate": 8.634526905381077e-06, "loss": 0.1586, "step": 3421 }, { "epoch": 1.06, "learning_rate": 8.634126825365074e-06, "loss": 0.1853, "step": 3422 }, { "epoch": 1.06, "learning_rate": 8.63372674534907e-06, "loss": 0.0544, "step": 3423 }, { "epoch": 1.06, "learning_rate": 8.633326665333067e-06, "loss": 0.2652, "step": 3424 }, { "epoch": 1.06, "learning_rate": 8.632926585317064e-06, "loss": 0.144, "step": 3425 }, { "epoch": 1.06, "learning_rate": 8.63252650530106e-06, "loss": 0.1625, "step": 3426 }, { "epoch": 1.06, "learning_rate": 8.632126425285057e-06, "loss": 0.1432, "step": 3427 }, { "epoch": 1.06, "learning_rate": 8.631726345269054e-06, "loss": 0.0744, "step": 3428 }, { "epoch": 1.06, "learning_rate": 8.63132626525305e-06, "loss": 0.0703, "step": 3429 }, { "epoch": 1.06, "learning_rate": 8.63092618523705e-06, "loss": 0.2775, "step": 3430 }, { "epoch": 1.06, "learning_rate": 8.630526105221046e-06, "loss": 0.2212, "step": 3431 }, { "epoch": 1.07, "learning_rate": 8.630126025205041e-06, "loss": 0.118, "step": 3432 }, { "epoch": 1.07, "learning_rate": 8.629725945189038e-06, "loss": 0.1895, "step": 3433 }, { "epoch": 1.07, "learning_rate": 8.629325865173034e-06, "loss": 0.0912, "step": 3434 }, { "epoch": 1.07, "learning_rate": 8.628925785157033e-06, "loss": 0.1776, "step": 3435 }, { "epoch": 1.07, "learning_rate": 8.62852570514103e-06, "loss": 0.1644, "step": 3436 }, { "epoch": 1.07, "learning_rate": 8.628125625125026e-06, "loss": 0.1855, "step": 3437 }, { "epoch": 1.07, "learning_rate": 8.627725545109023e-06, "loss": 0.0714, "step": 3438 }, { "epoch": 1.07, "learning_rate": 8.62732546509302e-06, "loss": 0.2408, "step": 3439 }, { "epoch": 1.07, "learning_rate": 8.626925385077016e-06, "loss": 0.1354, "step": 3440 }, { "epoch": 1.07, "learning_rate": 8.626525305061013e-06, "loss": 0.1542, "step": 3441 }, { "epoch": 1.07, "learning_rate": 8.62612522504501e-06, "loss": 0.3622, "step": 3442 }, { "epoch": 1.07, "learning_rate": 8.625725145029006e-06, "loss": 0.1432, "step": 3443 }, { "epoch": 1.07, "learning_rate": 8.625325065013004e-06, "loss": 0.1568, "step": 3444 }, { "epoch": 1.07, "learning_rate": 8.624924984997e-06, "loss": 0.2089, "step": 3445 }, { "epoch": 1.07, "learning_rate": 8.624524904980996e-06, "loss": 0.1252, "step": 3446 }, { "epoch": 1.07, "learning_rate": 8.624124824964993e-06, "loss": 0.4006, "step": 3447 }, { "epoch": 1.07, "learning_rate": 8.62372474494899e-06, "loss": 0.1492, "step": 3448 }, { "epoch": 1.07, "learning_rate": 8.623324664932988e-06, "loss": 0.0952, "step": 3449 }, { "epoch": 1.07, "learning_rate": 8.622924584916985e-06, "loss": 0.2175, "step": 3450 }, { "epoch": 1.07, "learning_rate": 8.622524504900981e-06, "loss": 0.283, "step": 3451 }, { "epoch": 1.07, "learning_rate": 8.622124424884978e-06, "loss": 0.1399, "step": 3452 }, { "epoch": 1.07, "learning_rate": 8.621724344868975e-06, "loss": 0.2711, "step": 3453 }, { "epoch": 1.07, "learning_rate": 8.621324264852971e-06, "loss": 0.1271, "step": 3454 }, { "epoch": 1.07, "learning_rate": 8.620924184836968e-06, "loss": 0.0912, "step": 3455 }, { "epoch": 1.07, "learning_rate": 8.620524104820965e-06, "loss": 0.2925, "step": 3456 }, { "epoch": 1.07, "learning_rate": 8.620124024804961e-06, "loss": 0.1302, "step": 3457 }, { "epoch": 1.07, "learning_rate": 8.619723944788958e-06, "loss": 0.1301, "step": 3458 }, { "epoch": 1.07, "learning_rate": 8.619323864772955e-06, "loss": 0.1696, "step": 3459 }, { "epoch": 1.07, "learning_rate": 8.618923784756951e-06, "loss": 0.1826, "step": 3460 }, { "epoch": 1.07, "learning_rate": 8.618523704740948e-06, "loss": 0.1166, "step": 3461 }, { "epoch": 1.07, "learning_rate": 8.618123624724945e-06, "loss": 0.1013, "step": 3462 }, { "epoch": 1.07, "learning_rate": 8.617723544708943e-06, "loss": 0.1026, "step": 3463 }, { "epoch": 1.08, "learning_rate": 8.61732346469294e-06, "loss": 0.0971, "step": 3464 }, { "epoch": 1.08, "learning_rate": 8.616923384676937e-06, "loss": 0.149, "step": 3465 }, { "epoch": 1.08, "learning_rate": 8.616523304660933e-06, "loss": 0.1057, "step": 3466 }, { "epoch": 1.08, "learning_rate": 8.61612322464493e-06, "loss": 0.2335, "step": 3467 }, { "epoch": 1.08, "learning_rate": 8.615723144628927e-06, "loss": 0.1217, "step": 3468 }, { "epoch": 1.08, "learning_rate": 8.615323064612923e-06, "loss": 0.1492, "step": 3469 }, { "epoch": 1.08, "learning_rate": 8.61492298459692e-06, "loss": 0.1477, "step": 3470 }, { "epoch": 1.08, "learning_rate": 8.614522904580917e-06, "loss": 0.1447, "step": 3471 }, { "epoch": 1.08, "learning_rate": 8.614122824564913e-06, "loss": 0.1351, "step": 3472 }, { "epoch": 1.08, "learning_rate": 8.61372274454891e-06, "loss": 0.1053, "step": 3473 }, { "epoch": 1.08, "learning_rate": 8.613322664532907e-06, "loss": 0.1481, "step": 3474 }, { "epoch": 1.08, "learning_rate": 8.612922584516903e-06, "loss": 0.1087, "step": 3475 }, { "epoch": 1.08, "learning_rate": 8.6125225045009e-06, "loss": 0.1848, "step": 3476 }, { "epoch": 1.08, "learning_rate": 8.612122424484899e-06, "loss": 0.2151, "step": 3477 }, { "epoch": 1.08, "learning_rate": 8.611722344468895e-06, "loss": 0.0854, "step": 3478 }, { "epoch": 1.08, "learning_rate": 8.611322264452892e-06, "loss": 0.0725, "step": 3479 }, { "epoch": 1.08, "learning_rate": 8.610922184436889e-06, "loss": 0.0907, "step": 3480 }, { "epoch": 1.08, "learning_rate": 8.610522104420885e-06, "loss": 0.187, "step": 3481 }, { "epoch": 1.08, "learning_rate": 8.610122024404882e-06, "loss": 0.2515, "step": 3482 }, { "epoch": 1.08, "learning_rate": 8.609721944388879e-06, "loss": 0.0973, "step": 3483 }, { "epoch": 1.08, "learning_rate": 8.609321864372875e-06, "loss": 0.1228, "step": 3484 }, { "epoch": 1.08, "learning_rate": 8.608921784356872e-06, "loss": 0.2304, "step": 3485 }, { "epoch": 1.08, "learning_rate": 8.608521704340869e-06, "loss": 0.1786, "step": 3486 }, { "epoch": 1.08, "learning_rate": 8.608121624324865e-06, "loss": 0.1963, "step": 3487 }, { "epoch": 1.08, "learning_rate": 8.607721544308862e-06, "loss": 0.1306, "step": 3488 }, { "epoch": 1.08, "learning_rate": 8.607321464292859e-06, "loss": 0.142, "step": 3489 }, { "epoch": 1.08, "learning_rate": 8.606921384276855e-06, "loss": 0.1501, "step": 3490 }, { "epoch": 1.08, "learning_rate": 8.606521304260854e-06, "loss": 0.1218, "step": 3491 }, { "epoch": 1.08, "learning_rate": 8.60612122424485e-06, "loss": 0.1377, "step": 3492 }, { "epoch": 1.08, "learning_rate": 8.605721144228847e-06, "loss": 0.1757, "step": 3493 }, { "epoch": 1.08, "learning_rate": 8.605321064212844e-06, "loss": 0.1885, "step": 3494 }, { "epoch": 1.08, "learning_rate": 8.604920984196839e-06, "loss": 0.1408, "step": 3495 }, { "epoch": 1.09, "learning_rate": 8.604520904180836e-06, "loss": 0.0952, "step": 3496 }, { "epoch": 1.09, "learning_rate": 8.604120824164834e-06, "loss": 0.1512, "step": 3497 }, { "epoch": 1.09, "learning_rate": 8.60372074414883e-06, "loss": 0.2378, "step": 3498 }, { "epoch": 1.09, "learning_rate": 8.603320664132827e-06, "loss": 0.0849, "step": 3499 }, { "epoch": 1.09, "learning_rate": 8.602920584116824e-06, "loss": 0.1361, "step": 3500 }, { "epoch": 1.09, "learning_rate": 8.60252050410082e-06, "loss": 0.1328, "step": 3501 }, { "epoch": 1.09, "learning_rate": 8.602120424084817e-06, "loss": 0.1529, "step": 3502 }, { "epoch": 1.09, "learning_rate": 8.601720344068814e-06, "loss": 0.1596, "step": 3503 }, { "epoch": 1.09, "learning_rate": 8.60132026405281e-06, "loss": 0.1844, "step": 3504 }, { "epoch": 1.09, "learning_rate": 8.600920184036809e-06, "loss": 0.0954, "step": 3505 }, { "epoch": 1.09, "learning_rate": 8.600520104020806e-06, "loss": 0.1067, "step": 3506 }, { "epoch": 1.09, "learning_rate": 8.600120024004803e-06, "loss": 0.1511, "step": 3507 }, { "epoch": 1.09, "learning_rate": 8.599719943988798e-06, "loss": 0.06, "step": 3508 }, { "epoch": 1.09, "learning_rate": 8.599319863972794e-06, "loss": 0.1897, "step": 3509 }, { "epoch": 1.09, "learning_rate": 8.598919783956791e-06, "loss": 0.1151, "step": 3510 }, { "epoch": 1.09, "learning_rate": 8.59851970394079e-06, "loss": 0.1398, "step": 3511 }, { "epoch": 1.09, "learning_rate": 8.598119623924786e-06, "loss": 0.1931, "step": 3512 }, { "epoch": 1.09, "learning_rate": 8.597719543908783e-06, "loss": 0.11, "step": 3513 }, { "epoch": 1.09, "learning_rate": 8.59731946389278e-06, "loss": 0.1449, "step": 3514 }, { "epoch": 1.09, "learning_rate": 8.596919383876776e-06, "loss": 0.2104, "step": 3515 }, { "epoch": 1.09, "learning_rate": 8.596519303860773e-06, "loss": 0.245, "step": 3516 }, { "epoch": 1.09, "learning_rate": 8.59611922384477e-06, "loss": 0.1625, "step": 3517 }, { "epoch": 1.09, "learning_rate": 8.595719143828766e-06, "loss": 0.1153, "step": 3518 }, { "epoch": 1.09, "learning_rate": 8.595319063812764e-06, "loss": 0.1992, "step": 3519 }, { "epoch": 1.09, "learning_rate": 8.594918983796761e-06, "loss": 0.2436, "step": 3520 }, { "epoch": 1.09, "learning_rate": 8.594518903780756e-06, "loss": 0.1006, "step": 3521 }, { "epoch": 1.09, "learning_rate": 8.594118823764753e-06, "loss": 0.099, "step": 3522 }, { "epoch": 1.09, "learning_rate": 8.59371874374875e-06, "loss": 0.2519, "step": 3523 }, { "epoch": 1.09, "learning_rate": 8.593318663732748e-06, "loss": 0.0907, "step": 3524 }, { "epoch": 1.09, "learning_rate": 8.592918583716745e-06, "loss": 0.2311, "step": 3525 }, { "epoch": 1.09, "learning_rate": 8.592518503700741e-06, "loss": 0.1394, "step": 3526 }, { "epoch": 1.09, "learning_rate": 8.592118423684738e-06, "loss": 0.1706, "step": 3527 }, { "epoch": 1.1, "learning_rate": 8.591718343668735e-06, "loss": 0.1388, "step": 3528 }, { "epoch": 1.1, "learning_rate": 8.591318263652731e-06, "loss": 0.1514, "step": 3529 }, { "epoch": 1.1, "learning_rate": 8.590918183636728e-06, "loss": 0.1803, "step": 3530 }, { "epoch": 1.1, "learning_rate": 8.590518103620725e-06, "loss": 0.206, "step": 3531 }, { "epoch": 1.1, "learning_rate": 8.590118023604721e-06, "loss": 0.1181, "step": 3532 }, { "epoch": 1.1, "learning_rate": 8.589717943588718e-06, "loss": 0.2285, "step": 3533 }, { "epoch": 1.1, "learning_rate": 8.589317863572715e-06, "loss": 0.106, "step": 3534 }, { "epoch": 1.1, "learning_rate": 8.588917783556711e-06, "loss": 0.2332, "step": 3535 }, { "epoch": 1.1, "learning_rate": 8.588517703540708e-06, "loss": 0.1315, "step": 3536 }, { "epoch": 1.1, "learning_rate": 8.588117623524705e-06, "loss": 0.1241, "step": 3537 }, { "epoch": 1.1, "learning_rate": 8.587717543508703e-06, "loss": 0.1361, "step": 3538 }, { "epoch": 1.1, "learning_rate": 8.5873174634927e-06, "loss": 0.0953, "step": 3539 }, { "epoch": 1.1, "learning_rate": 8.586917383476697e-06, "loss": 0.1326, "step": 3540 }, { "epoch": 1.1, "learning_rate": 8.586517303460693e-06, "loss": 0.1128, "step": 3541 }, { "epoch": 1.1, "learning_rate": 8.58611722344469e-06, "loss": 0.1282, "step": 3542 }, { "epoch": 1.1, "learning_rate": 8.585717143428687e-06, "loss": 0.226, "step": 3543 }, { "epoch": 1.1, "learning_rate": 8.585317063412683e-06, "loss": 0.1594, "step": 3544 }, { "epoch": 1.1, "learning_rate": 8.58491698339668e-06, "loss": 0.2094, "step": 3545 }, { "epoch": 1.1, "learning_rate": 8.584516903380677e-06, "loss": 0.1154, "step": 3546 }, { "epoch": 1.1, "learning_rate": 8.584116823364673e-06, "loss": 0.1956, "step": 3547 }, { "epoch": 1.1, "learning_rate": 8.58371674334867e-06, "loss": 0.1988, "step": 3548 }, { "epoch": 1.1, "learning_rate": 8.583316663332667e-06, "loss": 0.2784, "step": 3549 }, { "epoch": 1.1, "learning_rate": 8.582916583316663e-06, "loss": 0.22, "step": 3550 }, { "epoch": 1.1, "learning_rate": 8.58251650330066e-06, "loss": 0.1292, "step": 3551 }, { "epoch": 1.1, "learning_rate": 8.582116423284658e-06, "loss": 0.1298, "step": 3552 }, { "epoch": 1.1, "learning_rate": 8.581716343268655e-06, "loss": 0.216, "step": 3553 }, { "epoch": 1.1, "learning_rate": 8.581316263252652e-06, "loss": 0.1536, "step": 3554 }, { "epoch": 1.1, "learning_rate": 8.580916183236649e-06, "loss": 0.2378, "step": 3555 }, { "epoch": 1.1, "learning_rate": 8.580516103220645e-06, "loss": 0.1305, "step": 3556 }, { "epoch": 1.1, "learning_rate": 8.580116023204642e-06, "loss": 0.1027, "step": 3557 }, { "epoch": 1.1, "learning_rate": 8.579715943188639e-06, "loss": 0.1315, "step": 3558 }, { "epoch": 1.1, "learning_rate": 8.579315863172635e-06, "loss": 0.1723, "step": 3559 }, { "epoch": 1.1, "learning_rate": 8.578915783156632e-06, "loss": 0.2362, "step": 3560 }, { "epoch": 1.11, "learning_rate": 8.578515703140629e-06, "loss": 0.0935, "step": 3561 }, { "epoch": 1.11, "learning_rate": 8.578115623124625e-06, "loss": 0.2197, "step": 3562 }, { "epoch": 1.11, "learning_rate": 8.577715543108622e-06, "loss": 0.1575, "step": 3563 }, { "epoch": 1.11, "learning_rate": 8.577315463092619e-06, "loss": 0.1182, "step": 3564 }, { "epoch": 1.11, "learning_rate": 8.576915383076615e-06, "loss": 0.1303, "step": 3565 }, { "epoch": 1.11, "learning_rate": 8.576515303060614e-06, "loss": 0.2609, "step": 3566 }, { "epoch": 1.11, "learning_rate": 8.57611522304461e-06, "loss": 0.1508, "step": 3567 }, { "epoch": 1.11, "learning_rate": 8.575715143028607e-06, "loss": 0.2225, "step": 3568 }, { "epoch": 1.11, "learning_rate": 8.575315063012604e-06, "loss": 0.1653, "step": 3569 }, { "epoch": 1.11, "learning_rate": 8.5749149829966e-06, "loss": 0.2024, "step": 3570 }, { "epoch": 1.11, "learning_rate": 8.574514902980596e-06, "loss": 0.1277, "step": 3571 }, { "epoch": 1.11, "learning_rate": 8.574114822964594e-06, "loss": 0.1507, "step": 3572 }, { "epoch": 1.11, "learning_rate": 8.57371474294859e-06, "loss": 0.0914, "step": 3573 }, { "epoch": 1.11, "learning_rate": 8.573314662932587e-06, "loss": 0.0633, "step": 3574 }, { "epoch": 1.11, "learning_rate": 8.572914582916584e-06, "loss": 0.0626, "step": 3575 }, { "epoch": 1.11, "learning_rate": 8.57251450290058e-06, "loss": 0.1073, "step": 3576 }, { "epoch": 1.11, "learning_rate": 8.572114422884577e-06, "loss": 0.1932, "step": 3577 }, { "epoch": 1.11, "learning_rate": 8.571714342868574e-06, "loss": 0.1721, "step": 3578 }, { "epoch": 1.11, "learning_rate": 8.57131426285257e-06, "loss": 0.1444, "step": 3579 }, { "epoch": 1.11, "learning_rate": 8.570914182836569e-06, "loss": 0.2194, "step": 3580 }, { "epoch": 1.11, "learning_rate": 8.570514102820566e-06, "loss": 0.1087, "step": 3581 }, { "epoch": 1.11, "learning_rate": 8.570114022804562e-06, "loss": 0.1566, "step": 3582 }, { "epoch": 1.11, "learning_rate": 8.569713942788559e-06, "loss": 0.2041, "step": 3583 }, { "epoch": 1.11, "learning_rate": 8.569313862772554e-06, "loss": 0.1506, "step": 3584 }, { "epoch": 1.11, "learning_rate": 8.56891378275655e-06, "loss": 0.1784, "step": 3585 }, { "epoch": 1.11, "learning_rate": 8.56851370274055e-06, "loss": 0.1404, "step": 3586 }, { "epoch": 1.11, "learning_rate": 8.568113622724546e-06, "loss": 0.1778, "step": 3587 }, { "epoch": 1.11, "learning_rate": 8.567713542708543e-06, "loss": 0.1613, "step": 3588 }, { "epoch": 1.11, "learning_rate": 8.56731346269254e-06, "loss": 0.1144, "step": 3589 }, { "epoch": 1.11, "learning_rate": 8.566913382676536e-06, "loss": 0.1382, "step": 3590 }, { "epoch": 1.11, "learning_rate": 8.566513302660533e-06, "loss": 0.1116, "step": 3591 }, { "epoch": 1.11, "learning_rate": 8.56611322264453e-06, "loss": 0.1236, "step": 3592 }, { "epoch": 1.12, "learning_rate": 8.565713142628526e-06, "loss": 0.18, "step": 3593 }, { "epoch": 1.12, "learning_rate": 8.565313062612524e-06, "loss": 0.1964, "step": 3594 }, { "epoch": 1.12, "learning_rate": 8.564912982596521e-06, "loss": 0.1129, "step": 3595 }, { "epoch": 1.12, "learning_rate": 8.564512902580516e-06, "loss": 0.0979, "step": 3596 }, { "epoch": 1.12, "learning_rate": 8.564112822564513e-06, "loss": 0.1756, "step": 3597 }, { "epoch": 1.12, "learning_rate": 8.56371274254851e-06, "loss": 0.1492, "step": 3598 }, { "epoch": 1.12, "learning_rate": 8.563312662532506e-06, "loss": 0.1945, "step": 3599 }, { "epoch": 1.12, "learning_rate": 8.562912582516504e-06, "loss": 0.201, "step": 3600 }, { "epoch": 1.12, "learning_rate": 8.562512502500501e-06, "loss": 0.1256, "step": 3601 }, { "epoch": 1.12, "learning_rate": 8.562112422484498e-06, "loss": 0.0615, "step": 3602 }, { "epoch": 1.12, "learning_rate": 8.561712342468495e-06, "loss": 0.101, "step": 3603 }, { "epoch": 1.12, "learning_rate": 8.561312262452491e-06, "loss": 0.1211, "step": 3604 }, { "epoch": 1.12, "learning_rate": 8.560912182436488e-06, "loss": 0.0987, "step": 3605 }, { "epoch": 1.12, "learning_rate": 8.560512102420485e-06, "loss": 0.221, "step": 3606 }, { "epoch": 1.12, "learning_rate": 8.560112022404481e-06, "loss": 0.1873, "step": 3607 }, { "epoch": 1.12, "learning_rate": 8.55971194238848e-06, "loss": 0.3126, "step": 3608 }, { "epoch": 1.12, "learning_rate": 8.559311862372475e-06, "loss": 0.3422, "step": 3609 }, { "epoch": 1.12, "learning_rate": 8.558911782356471e-06, "loss": 0.0711, "step": 3610 }, { "epoch": 1.12, "learning_rate": 8.558511702340468e-06, "loss": 0.2291, "step": 3611 }, { "epoch": 1.12, "learning_rate": 8.558111622324465e-06, "loss": 0.1765, "step": 3612 }, { "epoch": 1.12, "learning_rate": 8.557711542308461e-06, "loss": 0.1203, "step": 3613 }, { "epoch": 1.12, "learning_rate": 8.55731146229246e-06, "loss": 0.1775, "step": 3614 }, { "epoch": 1.12, "learning_rate": 8.556911382276456e-06, "loss": 0.3727, "step": 3615 }, { "epoch": 1.12, "learning_rate": 8.556511302260453e-06, "loss": 0.2288, "step": 3616 }, { "epoch": 1.12, "learning_rate": 8.55611122224445e-06, "loss": 0.1077, "step": 3617 }, { "epoch": 1.12, "learning_rate": 8.555711142228447e-06, "loss": 0.1588, "step": 3618 }, { "epoch": 1.12, "learning_rate": 8.555311062212443e-06, "loss": 0.2768, "step": 3619 }, { "epoch": 1.12, "learning_rate": 8.55491098219644e-06, "loss": 0.1478, "step": 3620 }, { "epoch": 1.12, "learning_rate": 8.554510902180437e-06, "loss": 0.1679, "step": 3621 }, { "epoch": 1.12, "learning_rate": 8.554110822164433e-06, "loss": 0.1631, "step": 3622 }, { "epoch": 1.12, "learning_rate": 8.55371074214843e-06, "loss": 0.1127, "step": 3623 }, { "epoch": 1.12, "learning_rate": 8.553310662132427e-06, "loss": 0.1448, "step": 3624 }, { "epoch": 1.13, "learning_rate": 8.552910582116423e-06, "loss": 0.1726, "step": 3625 }, { "epoch": 1.13, "learning_rate": 8.55251050210042e-06, "loss": 0.1774, "step": 3626 }, { "epoch": 1.13, "learning_rate": 8.552110422084417e-06, "loss": 0.2117, "step": 3627 }, { "epoch": 1.13, "learning_rate": 8.551710342068415e-06, "loss": 0.1446, "step": 3628 }, { "epoch": 1.13, "learning_rate": 8.551310262052412e-06, "loss": 0.1212, "step": 3629 }, { "epoch": 1.13, "learning_rate": 8.550910182036408e-06, "loss": 0.1903, "step": 3630 }, { "epoch": 1.13, "learning_rate": 8.550510102020405e-06, "loss": 0.14, "step": 3631 }, { "epoch": 1.13, "learning_rate": 8.550110022004402e-06, "loss": 0.0683, "step": 3632 }, { "epoch": 1.13, "learning_rate": 8.549709941988399e-06, "loss": 0.1995, "step": 3633 }, { "epoch": 1.13, "learning_rate": 8.549309861972395e-06, "loss": 0.0822, "step": 3634 }, { "epoch": 1.13, "learning_rate": 8.548909781956392e-06, "loss": 0.1767, "step": 3635 }, { "epoch": 1.13, "learning_rate": 8.548509701940389e-06, "loss": 0.1615, "step": 3636 }, { "epoch": 1.13, "learning_rate": 8.548109621924385e-06, "loss": 0.3275, "step": 3637 }, { "epoch": 1.13, "learning_rate": 8.547709541908382e-06, "loss": 0.1326, "step": 3638 }, { "epoch": 1.13, "learning_rate": 8.547309461892379e-06, "loss": 0.1701, "step": 3639 }, { "epoch": 1.13, "learning_rate": 8.546909381876375e-06, "loss": 0.18, "step": 3640 }, { "epoch": 1.13, "learning_rate": 8.546509301860372e-06, "loss": 0.113, "step": 3641 }, { "epoch": 1.13, "learning_rate": 8.54610922184437e-06, "loss": 0.0945, "step": 3642 }, { "epoch": 1.13, "learning_rate": 8.545709141828367e-06, "loss": 0.2257, "step": 3643 }, { "epoch": 1.13, "learning_rate": 8.545309061812364e-06, "loss": 0.0993, "step": 3644 }, { "epoch": 1.13, "learning_rate": 8.54490898179636e-06, "loss": 0.1037, "step": 3645 }, { "epoch": 1.13, "learning_rate": 8.544508901780357e-06, "loss": 0.1334, "step": 3646 }, { "epoch": 1.13, "learning_rate": 8.544108821764354e-06, "loss": 0.243, "step": 3647 }, { "epoch": 1.13, "learning_rate": 8.54370874174835e-06, "loss": 0.1384, "step": 3648 }, { "epoch": 1.13, "learning_rate": 8.543308661732347e-06, "loss": 0.1717, "step": 3649 }, { "epoch": 1.13, "learning_rate": 8.542908581716344e-06, "loss": 0.2091, "step": 3650 }, { "epoch": 1.13, "learning_rate": 8.54250850170034e-06, "loss": 0.1324, "step": 3651 }, { "epoch": 1.13, "learning_rate": 8.542108421684337e-06, "loss": 0.1377, "step": 3652 }, { "epoch": 1.13, "learning_rate": 8.541708341668334e-06, "loss": 0.1082, "step": 3653 }, { "epoch": 1.13, "learning_rate": 8.54130826165233e-06, "loss": 0.1179, "step": 3654 }, { "epoch": 1.13, "learning_rate": 8.540908181636329e-06, "loss": 0.1948, "step": 3655 }, { "epoch": 1.13, "learning_rate": 8.540508101620326e-06, "loss": 0.2027, "step": 3656 }, { "epoch": 1.14, "learning_rate": 8.540108021604322e-06, "loss": 0.0968, "step": 3657 }, { "epoch": 1.14, "learning_rate": 8.539707941588319e-06, "loss": 0.0925, "step": 3658 }, { "epoch": 1.14, "learning_rate": 8.539307861572314e-06, "loss": 0.2137, "step": 3659 }, { "epoch": 1.14, "learning_rate": 8.53890778155631e-06, "loss": 0.0842, "step": 3660 }, { "epoch": 1.14, "learning_rate": 8.538507701540309e-06, "loss": 0.2311, "step": 3661 }, { "epoch": 1.14, "learning_rate": 8.538107621524306e-06, "loss": 0.2866, "step": 3662 }, { "epoch": 1.14, "learning_rate": 8.537707541508303e-06, "loss": 0.1435, "step": 3663 }, { "epoch": 1.14, "learning_rate": 8.5373074614923e-06, "loss": 0.166, "step": 3664 }, { "epoch": 1.14, "learning_rate": 8.536907381476296e-06, "loss": 0.348, "step": 3665 }, { "epoch": 1.14, "learning_rate": 8.536507301460293e-06, "loss": 0.2426, "step": 3666 }, { "epoch": 1.14, "learning_rate": 8.53610722144429e-06, "loss": 0.2407, "step": 3667 }, { "epoch": 1.14, "learning_rate": 8.535707141428286e-06, "loss": 0.2192, "step": 3668 }, { "epoch": 1.14, "learning_rate": 8.535307061412284e-06, "loss": 0.1138, "step": 3669 }, { "epoch": 1.14, "learning_rate": 8.534906981396281e-06, "loss": 0.1924, "step": 3670 }, { "epoch": 1.14, "learning_rate": 8.534506901380278e-06, "loss": 0.1955, "step": 3671 }, { "epoch": 1.14, "learning_rate": 8.534106821364273e-06, "loss": 0.1316, "step": 3672 }, { "epoch": 1.14, "learning_rate": 8.53370674134827e-06, "loss": 0.1763, "step": 3673 }, { "epoch": 1.14, "learning_rate": 8.533306661332266e-06, "loss": 0.1391, "step": 3674 }, { "epoch": 1.14, "learning_rate": 8.532906581316264e-06, "loss": 0.2178, "step": 3675 }, { "epoch": 1.14, "learning_rate": 8.532506501300261e-06, "loss": 0.1618, "step": 3676 }, { "epoch": 1.14, "learning_rate": 8.532106421284258e-06, "loss": 0.3204, "step": 3677 }, { "epoch": 1.14, "learning_rate": 8.531706341268254e-06, "loss": 0.1303, "step": 3678 }, { "epoch": 1.14, "learning_rate": 8.531306261252251e-06, "loss": 0.0997, "step": 3679 }, { "epoch": 1.14, "learning_rate": 8.530906181236248e-06, "loss": 0.1763, "step": 3680 }, { "epoch": 1.14, "learning_rate": 8.530506101220245e-06, "loss": 0.2835, "step": 3681 }, { "epoch": 1.14, "learning_rate": 8.530106021204241e-06, "loss": 0.1742, "step": 3682 }, { "epoch": 1.14, "learning_rate": 8.52970594118824e-06, "loss": 0.1379, "step": 3683 }, { "epoch": 1.14, "learning_rate": 8.529305861172236e-06, "loss": 0.1326, "step": 3684 }, { "epoch": 1.14, "learning_rate": 8.528905781156231e-06, "loss": 0.0639, "step": 3685 }, { "epoch": 1.14, "learning_rate": 8.528505701140228e-06, "loss": 0.1477, "step": 3686 }, { "epoch": 1.14, "learning_rate": 8.528105621124225e-06, "loss": 0.0893, "step": 3687 }, { "epoch": 1.14, "learning_rate": 8.527705541108221e-06, "loss": 0.134, "step": 3688 }, { "epoch": 1.14, "learning_rate": 8.52730546109222e-06, "loss": 0.217, "step": 3689 }, { "epoch": 1.15, "learning_rate": 8.526905381076216e-06, "loss": 0.207, "step": 3690 }, { "epoch": 1.15, "learning_rate": 8.526505301060213e-06, "loss": 0.1589, "step": 3691 }, { "epoch": 1.15, "learning_rate": 8.52610522104421e-06, "loss": 0.1742, "step": 3692 }, { "epoch": 1.15, "learning_rate": 8.525705141028206e-06, "loss": 0.1121, "step": 3693 }, { "epoch": 1.15, "learning_rate": 8.525305061012203e-06, "loss": 0.201, "step": 3694 }, { "epoch": 1.15, "learning_rate": 8.5249049809962e-06, "loss": 0.3621, "step": 3695 }, { "epoch": 1.15, "learning_rate": 8.524504900980197e-06, "loss": 0.399, "step": 3696 }, { "epoch": 1.15, "learning_rate": 8.524104820964193e-06, "loss": 0.2292, "step": 3697 }, { "epoch": 1.15, "learning_rate": 8.52370474094819e-06, "loss": 0.1728, "step": 3698 }, { "epoch": 1.15, "learning_rate": 8.523304660932187e-06, "loss": 0.0857, "step": 3699 }, { "epoch": 1.15, "learning_rate": 8.522904580916183e-06, "loss": 0.2244, "step": 3700 }, { "epoch": 1.15, "learning_rate": 8.52250450090018e-06, "loss": 0.1246, "step": 3701 }, { "epoch": 1.15, "learning_rate": 8.522104420884177e-06, "loss": 0.1446, "step": 3702 }, { "epoch": 1.15, "learning_rate": 8.521704340868175e-06, "loss": 0.1358, "step": 3703 }, { "epoch": 1.15, "learning_rate": 8.521304260852172e-06, "loss": 0.2606, "step": 3704 }, { "epoch": 1.15, "learning_rate": 8.520904180836168e-06, "loss": 0.1575, "step": 3705 }, { "epoch": 1.15, "learning_rate": 8.520504100820165e-06, "loss": 0.1743, "step": 3706 }, { "epoch": 1.15, "learning_rate": 8.520104020804162e-06, "loss": 0.183, "step": 3707 }, { "epoch": 1.15, "learning_rate": 8.519703940788158e-06, "loss": 0.2695, "step": 3708 }, { "epoch": 1.15, "learning_rate": 8.519303860772155e-06, "loss": 0.1733, "step": 3709 }, { "epoch": 1.15, "learning_rate": 8.518903780756152e-06, "loss": 0.2673, "step": 3710 }, { "epoch": 1.15, "learning_rate": 8.518503700740149e-06, "loss": 0.1664, "step": 3711 }, { "epoch": 1.15, "learning_rate": 8.518103620724145e-06, "loss": 0.1301, "step": 3712 }, { "epoch": 1.15, "learning_rate": 8.517703540708142e-06, "loss": 0.1978, "step": 3713 }, { "epoch": 1.15, "learning_rate": 8.517303460692139e-06, "loss": 0.1901, "step": 3714 }, { "epoch": 1.15, "learning_rate": 8.516903380676135e-06, "loss": 0.2333, "step": 3715 }, { "epoch": 1.15, "learning_rate": 8.516503300660132e-06, "loss": 0.1894, "step": 3716 }, { "epoch": 1.15, "learning_rate": 8.51610322064413e-06, "loss": 0.1431, "step": 3717 }, { "epoch": 1.15, "learning_rate": 8.515703140628127e-06, "loss": 0.1064, "step": 3718 }, { "epoch": 1.15, "learning_rate": 8.515303060612124e-06, "loss": 0.1528, "step": 3719 }, { "epoch": 1.15, "learning_rate": 8.51490298059612e-06, "loss": 0.1067, "step": 3720 }, { "epoch": 1.15, "learning_rate": 8.514502900580117e-06, "loss": 0.139, "step": 3721 }, { "epoch": 1.16, "learning_rate": 8.514102820564114e-06, "loss": 0.1679, "step": 3722 }, { "epoch": 1.16, "learning_rate": 8.51370274054811e-06, "loss": 0.1061, "step": 3723 }, { "epoch": 1.16, "learning_rate": 8.513302660532107e-06, "loss": 0.1594, "step": 3724 }, { "epoch": 1.16, "learning_rate": 8.512902580516104e-06, "loss": 0.1703, "step": 3725 }, { "epoch": 1.16, "learning_rate": 8.5125025005001e-06, "loss": 0.094, "step": 3726 }, { "epoch": 1.16, "learning_rate": 8.512102420484097e-06, "loss": 0.1128, "step": 3727 }, { "epoch": 1.16, "learning_rate": 8.511702340468094e-06, "loss": 0.2196, "step": 3728 }, { "epoch": 1.16, "learning_rate": 8.51130226045209e-06, "loss": 0.1284, "step": 3729 }, { "epoch": 1.16, "learning_rate": 8.510902180436087e-06, "loss": 0.1825, "step": 3730 }, { "epoch": 1.16, "learning_rate": 8.510502100420086e-06, "loss": 0.276, "step": 3731 }, { "epoch": 1.16, "learning_rate": 8.510102020404082e-06, "loss": 0.2031, "step": 3732 }, { "epoch": 1.16, "learning_rate": 8.509701940388079e-06, "loss": 0.1325, "step": 3733 }, { "epoch": 1.16, "learning_rate": 8.509301860372076e-06, "loss": 0.2421, "step": 3734 }, { "epoch": 1.16, "learning_rate": 8.50890178035607e-06, "loss": 0.1264, "step": 3735 }, { "epoch": 1.16, "learning_rate": 8.508501700340069e-06, "loss": 0.1436, "step": 3736 }, { "epoch": 1.16, "learning_rate": 8.508101620324066e-06, "loss": 0.1688, "step": 3737 }, { "epoch": 1.16, "learning_rate": 8.507701540308062e-06, "loss": 0.3174, "step": 3738 }, { "epoch": 1.16, "learning_rate": 8.507301460292059e-06, "loss": 0.1719, "step": 3739 }, { "epoch": 1.16, "learning_rate": 8.506901380276056e-06, "loss": 0.1943, "step": 3740 }, { "epoch": 1.16, "learning_rate": 8.506501300260053e-06, "loss": 0.2119, "step": 3741 }, { "epoch": 1.16, "learning_rate": 8.50610122024405e-06, "loss": 0.1629, "step": 3742 }, { "epoch": 1.16, "learning_rate": 8.505701140228046e-06, "loss": 0.1607, "step": 3743 }, { "epoch": 1.16, "learning_rate": 8.505301060212043e-06, "loss": 0.2462, "step": 3744 }, { "epoch": 1.16, "learning_rate": 8.504900980196041e-06, "loss": 0.1978, "step": 3745 }, { "epoch": 1.16, "learning_rate": 8.504500900180038e-06, "loss": 0.1077, "step": 3746 }, { "epoch": 1.16, "learning_rate": 8.504100820164034e-06, "loss": 0.1838, "step": 3747 }, { "epoch": 1.16, "learning_rate": 8.50370074014803e-06, "loss": 0.3378, "step": 3748 }, { "epoch": 1.16, "learning_rate": 8.503300660132026e-06, "loss": 0.1498, "step": 3749 }, { "epoch": 1.16, "learning_rate": 8.502900580116024e-06, "loss": 0.1586, "step": 3750 }, { "epoch": 1.16, "learning_rate": 8.502500500100021e-06, "loss": 0.1484, "step": 3751 }, { "epoch": 1.16, "learning_rate": 8.502100420084018e-06, "loss": 0.2138, "step": 3752 }, { "epoch": 1.16, "learning_rate": 8.501700340068014e-06, "loss": 0.1125, "step": 3753 }, { "epoch": 1.17, "learning_rate": 8.501300260052011e-06, "loss": 0.1741, "step": 3754 }, { "epoch": 1.17, "learning_rate": 8.500900180036008e-06, "loss": 0.1491, "step": 3755 }, { "epoch": 1.17, "learning_rate": 8.500500100020004e-06, "loss": 0.1694, "step": 3756 }, { "epoch": 1.17, "learning_rate": 8.500100020004001e-06, "loss": 0.1775, "step": 3757 }, { "epoch": 1.17, "learning_rate": 8.499699939987998e-06, "loss": 0.1231, "step": 3758 }, { "epoch": 1.17, "learning_rate": 8.499299859971996e-06, "loss": 0.0743, "step": 3759 }, { "epoch": 1.17, "learning_rate": 8.498899779955991e-06, "loss": 0.204, "step": 3760 }, { "epoch": 1.17, "learning_rate": 8.498499699939988e-06, "loss": 0.1167, "step": 3761 }, { "epoch": 1.17, "learning_rate": 8.498099619923985e-06, "loss": 0.1606, "step": 3762 }, { "epoch": 1.17, "learning_rate": 8.497699539907981e-06, "loss": 0.142, "step": 3763 }, { "epoch": 1.17, "learning_rate": 8.49729945989198e-06, "loss": 0.141, "step": 3764 }, { "epoch": 1.17, "learning_rate": 8.496899379875976e-06, "loss": 0.1956, "step": 3765 }, { "epoch": 1.17, "learning_rate": 8.496499299859973e-06, "loss": 0.1398, "step": 3766 }, { "epoch": 1.17, "learning_rate": 8.49609921984397e-06, "loss": 0.219, "step": 3767 }, { "epoch": 1.17, "learning_rate": 8.495699139827966e-06, "loss": 0.1738, "step": 3768 }, { "epoch": 1.17, "learning_rate": 8.495299059811963e-06, "loss": 0.1719, "step": 3769 }, { "epoch": 1.17, "learning_rate": 8.49489897979596e-06, "loss": 0.1612, "step": 3770 }, { "epoch": 1.17, "learning_rate": 8.494498899779956e-06, "loss": 0.1983, "step": 3771 }, { "epoch": 1.17, "learning_rate": 8.494098819763953e-06, "loss": 0.1292, "step": 3772 }, { "epoch": 1.17, "learning_rate": 8.49369873974795e-06, "loss": 0.144, "step": 3773 }, { "epoch": 1.17, "learning_rate": 8.493298659731947e-06, "loss": 0.0803, "step": 3774 }, { "epoch": 1.17, "learning_rate": 8.492898579715943e-06, "loss": 0.1941, "step": 3775 }, { "epoch": 1.17, "learning_rate": 8.49249849969994e-06, "loss": 0.2429, "step": 3776 }, { "epoch": 1.17, "learning_rate": 8.492098419683937e-06, "loss": 0.2146, "step": 3777 }, { "epoch": 1.17, "learning_rate": 8.491698339667935e-06, "loss": 0.2183, "step": 3778 }, { "epoch": 1.17, "learning_rate": 8.491298259651932e-06, "loss": 0.3072, "step": 3779 }, { "epoch": 1.17, "learning_rate": 8.490898179635928e-06, "loss": 0.1362, "step": 3780 }, { "epoch": 1.17, "learning_rate": 8.490498099619925e-06, "loss": 0.1238, "step": 3781 }, { "epoch": 1.17, "learning_rate": 8.490098019603922e-06, "loss": 0.1107, "step": 3782 }, { "epoch": 1.17, "learning_rate": 8.489697939587918e-06, "loss": 0.2251, "step": 3783 }, { "epoch": 1.17, "learning_rate": 8.489297859571915e-06, "loss": 0.3134, "step": 3784 }, { "epoch": 1.17, "learning_rate": 8.488897779555912e-06, "loss": 0.1257, "step": 3785 }, { "epoch": 1.18, "learning_rate": 8.488497699539908e-06, "loss": 0.2446, "step": 3786 }, { "epoch": 1.18, "learning_rate": 8.488097619523905e-06, "loss": 0.1465, "step": 3787 }, { "epoch": 1.18, "learning_rate": 8.487697539507902e-06, "loss": 0.1553, "step": 3788 }, { "epoch": 1.18, "learning_rate": 8.487297459491899e-06, "loss": 0.1791, "step": 3789 }, { "epoch": 1.18, "learning_rate": 8.486897379475895e-06, "loss": 0.2045, "step": 3790 }, { "epoch": 1.18, "learning_rate": 8.486497299459892e-06, "loss": 0.26, "step": 3791 }, { "epoch": 1.18, "learning_rate": 8.48609721944389e-06, "loss": 0.1867, "step": 3792 }, { "epoch": 1.18, "learning_rate": 8.485697139427887e-06, "loss": 0.1529, "step": 3793 }, { "epoch": 1.18, "learning_rate": 8.485297059411884e-06, "loss": 0.086, "step": 3794 }, { "epoch": 1.18, "learning_rate": 8.48489697939588e-06, "loss": 0.2687, "step": 3795 }, { "epoch": 1.18, "learning_rate": 8.484496899379877e-06, "loss": 0.0994, "step": 3796 }, { "epoch": 1.18, "learning_rate": 8.484096819363874e-06, "loss": 0.154, "step": 3797 }, { "epoch": 1.18, "learning_rate": 8.48369673934787e-06, "loss": 0.102, "step": 3798 }, { "epoch": 1.18, "learning_rate": 8.483296659331867e-06, "loss": 0.0868, "step": 3799 }, { "epoch": 1.18, "learning_rate": 8.482896579315864e-06, "loss": 0.1029, "step": 3800 }, { "epoch": 1.18, "learning_rate": 8.48249649929986e-06, "loss": 0.1655, "step": 3801 }, { "epoch": 1.18, "learning_rate": 8.482096419283857e-06, "loss": 0.1541, "step": 3802 }, { "epoch": 1.18, "learning_rate": 8.481696339267854e-06, "loss": 0.162, "step": 3803 }, { "epoch": 1.18, "learning_rate": 8.48129625925185e-06, "loss": 0.1362, "step": 3804 }, { "epoch": 1.18, "learning_rate": 8.480896179235847e-06, "loss": 0.1655, "step": 3805 }, { "epoch": 1.18, "learning_rate": 8.480496099219846e-06, "loss": 0.1406, "step": 3806 }, { "epoch": 1.18, "learning_rate": 8.480096019203842e-06, "loss": 0.1197, "step": 3807 }, { "epoch": 1.18, "learning_rate": 8.479695939187839e-06, "loss": 0.1447, "step": 3808 }, { "epoch": 1.18, "learning_rate": 8.479295859171836e-06, "loss": 0.1355, "step": 3809 }, { "epoch": 1.18, "learning_rate": 8.478895779155832e-06, "loss": 0.1589, "step": 3810 }, { "epoch": 1.18, "learning_rate": 8.478495699139827e-06, "loss": 0.1644, "step": 3811 }, { "epoch": 1.18, "learning_rate": 8.478095619123826e-06, "loss": 0.1623, "step": 3812 }, { "epoch": 1.18, "learning_rate": 8.477695539107822e-06, "loss": 0.1415, "step": 3813 }, { "epoch": 1.18, "learning_rate": 8.477295459091819e-06, "loss": 0.0762, "step": 3814 }, { "epoch": 1.18, "learning_rate": 8.476895379075816e-06, "loss": 0.1557, "step": 3815 }, { "epoch": 1.18, "learning_rate": 8.476495299059812e-06, "loss": 0.2955, "step": 3816 }, { "epoch": 1.18, "learning_rate": 8.476095219043809e-06, "loss": 0.1205, "step": 3817 }, { "epoch": 1.19, "learning_rate": 8.475695139027806e-06, "loss": 0.1096, "step": 3818 }, { "epoch": 1.19, "learning_rate": 8.475295059011803e-06, "loss": 0.1249, "step": 3819 }, { "epoch": 1.19, "learning_rate": 8.474894978995801e-06, "loss": 0.2861, "step": 3820 }, { "epoch": 1.19, "learning_rate": 8.474494898979798e-06, "loss": 0.1113, "step": 3821 }, { "epoch": 1.19, "learning_rate": 8.474094818963794e-06, "loss": 0.1101, "step": 3822 }, { "epoch": 1.19, "learning_rate": 8.47369473894779e-06, "loss": 0.1784, "step": 3823 }, { "epoch": 1.19, "learning_rate": 8.473294658931786e-06, "loss": 0.2074, "step": 3824 }, { "epoch": 1.19, "learning_rate": 8.472894578915784e-06, "loss": 0.121, "step": 3825 }, { "epoch": 1.19, "learning_rate": 8.472494498899781e-06, "loss": 0.1503, "step": 3826 }, { "epoch": 1.19, "learning_rate": 8.472094418883778e-06, "loss": 0.0761, "step": 3827 }, { "epoch": 1.19, "learning_rate": 8.471694338867774e-06, "loss": 0.1153, "step": 3828 }, { "epoch": 1.19, "learning_rate": 8.471294258851771e-06, "loss": 0.1771, "step": 3829 }, { "epoch": 1.19, "learning_rate": 8.470894178835768e-06, "loss": 0.1813, "step": 3830 }, { "epoch": 1.19, "learning_rate": 8.470494098819764e-06, "loss": 0.1671, "step": 3831 }, { "epoch": 1.19, "learning_rate": 8.470094018803761e-06, "loss": 0.0987, "step": 3832 }, { "epoch": 1.19, "learning_rate": 8.469693938787758e-06, "loss": 0.0726, "step": 3833 }, { "epoch": 1.19, "learning_rate": 8.469293858771756e-06, "loss": 0.1275, "step": 3834 }, { "epoch": 1.19, "learning_rate": 8.468893778755753e-06, "loss": 0.1665, "step": 3835 }, { "epoch": 1.19, "learning_rate": 8.468493698739748e-06, "loss": 0.1455, "step": 3836 }, { "epoch": 1.19, "learning_rate": 8.468093618723745e-06, "loss": 0.142, "step": 3837 }, { "epoch": 1.19, "learning_rate": 8.467693538707741e-06, "loss": 0.2499, "step": 3838 }, { "epoch": 1.19, "learning_rate": 8.46729345869174e-06, "loss": 0.1217, "step": 3839 }, { "epoch": 1.19, "learning_rate": 8.466893378675736e-06, "loss": 0.2314, "step": 3840 }, { "epoch": 1.19, "learning_rate": 8.466493298659733e-06, "loss": 0.1878, "step": 3841 }, { "epoch": 1.19, "learning_rate": 8.46609321864373e-06, "loss": 0.2149, "step": 3842 }, { "epoch": 1.19, "learning_rate": 8.465693138627726e-06, "loss": 0.1948, "step": 3843 }, { "epoch": 1.19, "learning_rate": 8.465293058611723e-06, "loss": 0.2115, "step": 3844 }, { "epoch": 1.19, "learning_rate": 8.46489297859572e-06, "loss": 0.3095, "step": 3845 }, { "epoch": 1.19, "learning_rate": 8.464492898579716e-06, "loss": 0.154, "step": 3846 }, { "epoch": 1.19, "learning_rate": 8.464092818563713e-06, "loss": 0.287, "step": 3847 }, { "epoch": 1.19, "learning_rate": 8.463692738547711e-06, "loss": 0.2649, "step": 3848 }, { "epoch": 1.19, "learning_rate": 8.463292658531706e-06, "loss": 0.1161, "step": 3849 }, { "epoch": 1.19, "learning_rate": 8.462892578515703e-06, "loss": 0.2431, "step": 3850 }, { "epoch": 1.2, "learning_rate": 8.4624924984997e-06, "loss": 0.1299, "step": 3851 }, { "epoch": 1.2, "learning_rate": 8.462092418483697e-06, "loss": 0.1365, "step": 3852 }, { "epoch": 1.2, "learning_rate": 8.461692338467695e-06, "loss": 0.1424, "step": 3853 }, { "epoch": 1.2, "learning_rate": 8.461292258451692e-06, "loss": 0.2084, "step": 3854 }, { "epoch": 1.2, "learning_rate": 8.460892178435688e-06, "loss": 0.1289, "step": 3855 }, { "epoch": 1.2, "learning_rate": 8.460492098419685e-06, "loss": 0.0761, "step": 3856 }, { "epoch": 1.2, "learning_rate": 8.460092018403682e-06, "loss": 0.1301, "step": 3857 }, { "epoch": 1.2, "learning_rate": 8.459691938387678e-06, "loss": 0.0821, "step": 3858 }, { "epoch": 1.2, "learning_rate": 8.459291858371675e-06, "loss": 0.1224, "step": 3859 }, { "epoch": 1.2, "learning_rate": 8.458891778355672e-06, "loss": 0.1188, "step": 3860 }, { "epoch": 1.2, "learning_rate": 8.458491698339668e-06, "loss": 0.0988, "step": 3861 }, { "epoch": 1.2, "learning_rate": 8.458091618323665e-06, "loss": 0.1613, "step": 3862 }, { "epoch": 1.2, "learning_rate": 8.457691538307662e-06, "loss": 0.1458, "step": 3863 }, { "epoch": 1.2, "learning_rate": 8.457291458291658e-06, "loss": 0.1482, "step": 3864 }, { "epoch": 1.2, "learning_rate": 8.456891378275655e-06, "loss": 0.1182, "step": 3865 }, { "epoch": 1.2, "learning_rate": 8.456491298259652e-06, "loss": 0.1541, "step": 3866 }, { "epoch": 1.2, "learning_rate": 8.45609121824365e-06, "loss": 0.0772, "step": 3867 }, { "epoch": 1.2, "learning_rate": 8.455691138227647e-06, "loss": 0.1467, "step": 3868 }, { "epoch": 1.2, "learning_rate": 8.455291058211644e-06, "loss": 0.1292, "step": 3869 }, { "epoch": 1.2, "learning_rate": 8.45489097819564e-06, "loss": 0.109, "step": 3870 }, { "epoch": 1.2, "learning_rate": 8.454490898179637e-06, "loss": 0.0802, "step": 3871 }, { "epoch": 1.2, "learning_rate": 8.454090818163634e-06, "loss": 0.0802, "step": 3872 }, { "epoch": 1.2, "learning_rate": 8.45369073814763e-06, "loss": 0.1498, "step": 3873 }, { "epoch": 1.2, "learning_rate": 8.453290658131627e-06, "loss": 0.1159, "step": 3874 }, { "epoch": 1.2, "learning_rate": 8.452890578115624e-06, "loss": 0.302, "step": 3875 }, { "epoch": 1.2, "learning_rate": 8.45249049809962e-06, "loss": 0.162, "step": 3876 }, { "epoch": 1.2, "learning_rate": 8.452090418083617e-06, "loss": 0.267, "step": 3877 }, { "epoch": 1.2, "learning_rate": 8.451690338067614e-06, "loss": 0.0732, "step": 3878 }, { "epoch": 1.2, "learning_rate": 8.45129025805161e-06, "loss": 0.1557, "step": 3879 }, { "epoch": 1.2, "learning_rate": 8.450890178035607e-06, "loss": 0.1566, "step": 3880 }, { "epoch": 1.2, "learning_rate": 8.450490098019606e-06, "loss": 0.1267, "step": 3881 }, { "epoch": 1.2, "learning_rate": 8.450090018003602e-06, "loss": 0.3018, "step": 3882 }, { "epoch": 1.21, "learning_rate": 8.449689937987599e-06, "loss": 0.2193, "step": 3883 }, { "epoch": 1.21, "learning_rate": 8.449289857971596e-06, "loss": 0.098, "step": 3884 }, { "epoch": 1.21, "learning_rate": 8.448889777955592e-06, "loss": 0.0751, "step": 3885 }, { "epoch": 1.21, "learning_rate": 8.448489697939587e-06, "loss": 0.2151, "step": 3886 }, { "epoch": 1.21, "learning_rate": 8.448089617923586e-06, "loss": 0.1154, "step": 3887 }, { "epoch": 1.21, "learning_rate": 8.447689537907582e-06, "loss": 0.1414, "step": 3888 }, { "epoch": 1.21, "learning_rate": 8.447289457891579e-06, "loss": 0.061, "step": 3889 }, { "epoch": 1.21, "learning_rate": 8.446889377875576e-06, "loss": 0.0833, "step": 3890 }, { "epoch": 1.21, "learning_rate": 8.446489297859572e-06, "loss": 0.0769, "step": 3891 }, { "epoch": 1.21, "learning_rate": 8.446089217843569e-06, "loss": 0.1511, "step": 3892 }, { "epoch": 1.21, "learning_rate": 8.445689137827566e-06, "loss": 0.0883, "step": 3893 }, { "epoch": 1.21, "learning_rate": 8.445289057811562e-06, "loss": 0.1678, "step": 3894 }, { "epoch": 1.21, "learning_rate": 8.44488897779556e-06, "loss": 0.1362, "step": 3895 }, { "epoch": 1.21, "learning_rate": 8.444488897779558e-06, "loss": 0.1421, "step": 3896 }, { "epoch": 1.21, "learning_rate": 8.444088817763554e-06, "loss": 0.2775, "step": 3897 }, { "epoch": 1.21, "learning_rate": 8.443688737747551e-06, "loss": 0.1635, "step": 3898 }, { "epoch": 1.21, "learning_rate": 8.443288657731546e-06, "loss": 0.121, "step": 3899 }, { "epoch": 1.21, "learning_rate": 8.442888577715543e-06, "loss": 0.106, "step": 3900 }, { "epoch": 1.21, "learning_rate": 8.442488497699541e-06, "loss": 0.1361, "step": 3901 }, { "epoch": 1.21, "learning_rate": 8.442088417683538e-06, "loss": 0.2793, "step": 3902 }, { "epoch": 1.21, "learning_rate": 8.441688337667534e-06, "loss": 0.3746, "step": 3903 }, { "epoch": 1.21, "learning_rate": 8.441288257651531e-06, "loss": 0.0888, "step": 3904 }, { "epoch": 1.21, "learning_rate": 8.440888177635528e-06, "loss": 0.156, "step": 3905 }, { "epoch": 1.21, "learning_rate": 8.440488097619524e-06, "loss": 0.1596, "step": 3906 }, { "epoch": 1.21, "learning_rate": 8.440088017603521e-06, "loss": 0.1484, "step": 3907 }, { "epoch": 1.21, "learning_rate": 8.439687937587518e-06, "loss": 0.1587, "step": 3908 }, { "epoch": 1.21, "learning_rate": 8.439287857571516e-06, "loss": 0.1627, "step": 3909 }, { "epoch": 1.21, "learning_rate": 8.438887777555513e-06, "loss": 0.196, "step": 3910 }, { "epoch": 1.21, "learning_rate": 8.43848769753951e-06, "loss": 0.2432, "step": 3911 }, { "epoch": 1.21, "learning_rate": 8.438087617523504e-06, "loss": 0.0907, "step": 3912 }, { "epoch": 1.21, "learning_rate": 8.437687537507501e-06, "loss": 0.2852, "step": 3913 }, { "epoch": 1.21, "learning_rate": 8.437287457491498e-06, "loss": 0.1218, "step": 3914 }, { "epoch": 1.22, "learning_rate": 8.436887377475496e-06, "loss": 0.1603, "step": 3915 }, { "epoch": 1.22, "learning_rate": 8.436487297459493e-06, "loss": 0.1368, "step": 3916 }, { "epoch": 1.22, "learning_rate": 8.43608721744349e-06, "loss": 0.2991, "step": 3917 }, { "epoch": 1.22, "learning_rate": 8.435687137427486e-06, "loss": 0.2016, "step": 3918 }, { "epoch": 1.22, "learning_rate": 8.435287057411483e-06, "loss": 0.2382, "step": 3919 }, { "epoch": 1.22, "learning_rate": 8.43488697739548e-06, "loss": 0.1415, "step": 3920 }, { "epoch": 1.22, "learning_rate": 8.434486897379476e-06, "loss": 0.1227, "step": 3921 }, { "epoch": 1.22, "learning_rate": 8.434086817363473e-06, "loss": 0.276, "step": 3922 }, { "epoch": 1.22, "learning_rate": 8.433686737347471e-06, "loss": 0.1774, "step": 3923 }, { "epoch": 1.22, "learning_rate": 8.433286657331466e-06, "loss": 0.1536, "step": 3924 }, { "epoch": 1.22, "learning_rate": 8.432886577315463e-06, "loss": 0.1821, "step": 3925 }, { "epoch": 1.22, "learning_rate": 8.43248649729946e-06, "loss": 0.1057, "step": 3926 }, { "epoch": 1.22, "learning_rate": 8.432086417283456e-06, "loss": 0.1225, "step": 3927 }, { "epoch": 1.22, "learning_rate": 8.431686337267453e-06, "loss": 0.1484, "step": 3928 }, { "epoch": 1.22, "learning_rate": 8.431286257251452e-06, "loss": 0.1975, "step": 3929 }, { "epoch": 1.22, "learning_rate": 8.430886177235448e-06, "loss": 0.2062, "step": 3930 }, { "epoch": 1.22, "learning_rate": 8.430486097219445e-06, "loss": 0.1917, "step": 3931 }, { "epoch": 1.22, "learning_rate": 8.430086017203442e-06, "loss": 0.1629, "step": 3932 }, { "epoch": 1.22, "learning_rate": 8.429685937187438e-06, "loss": 0.0503, "step": 3933 }, { "epoch": 1.22, "learning_rate": 8.429285857171435e-06, "loss": 0.1921, "step": 3934 }, { "epoch": 1.22, "learning_rate": 8.428885777155432e-06, "loss": 0.2044, "step": 3935 }, { "epoch": 1.22, "learning_rate": 8.428485697139428e-06, "loss": 0.0742, "step": 3936 }, { "epoch": 1.22, "learning_rate": 8.428085617123425e-06, "loss": 0.1427, "step": 3937 }, { "epoch": 1.22, "learning_rate": 8.427685537107422e-06, "loss": 0.1712, "step": 3938 }, { "epoch": 1.22, "learning_rate": 8.427285457091418e-06, "loss": 0.1169, "step": 3939 }, { "epoch": 1.22, "learning_rate": 8.426885377075415e-06, "loss": 0.1234, "step": 3940 }, { "epoch": 1.22, "learning_rate": 8.426485297059412e-06, "loss": 0.1158, "step": 3941 }, { "epoch": 1.22, "learning_rate": 8.426085217043408e-06, "loss": 0.0865, "step": 3942 }, { "epoch": 1.22, "learning_rate": 8.425685137027407e-06, "loss": 0.1499, "step": 3943 }, { "epoch": 1.22, "learning_rate": 8.425285057011404e-06, "loss": 0.1242, "step": 3944 }, { "epoch": 1.22, "learning_rate": 8.4248849769954e-06, "loss": 0.1396, "step": 3945 }, { "epoch": 1.22, "learning_rate": 8.424484896979397e-06, "loss": 0.0968, "step": 3946 }, { "epoch": 1.23, "learning_rate": 8.424084816963394e-06, "loss": 0.1849, "step": 3947 }, { "epoch": 1.23, "learning_rate": 8.42368473694739e-06, "loss": 0.1681, "step": 3948 }, { "epoch": 1.23, "learning_rate": 8.423284656931387e-06, "loss": 0.1034, "step": 3949 }, { "epoch": 1.23, "learning_rate": 8.422884576915384e-06, "loss": 0.1184, "step": 3950 }, { "epoch": 1.23, "learning_rate": 8.42248449689938e-06, "loss": 0.1678, "step": 3951 }, { "epoch": 1.23, "learning_rate": 8.422084416883377e-06, "loss": 0.0835, "step": 3952 }, { "epoch": 1.23, "learning_rate": 8.421684336867374e-06, "loss": 0.153, "step": 3953 }, { "epoch": 1.23, "learning_rate": 8.42128425685137e-06, "loss": 0.1142, "step": 3954 }, { "epoch": 1.23, "learning_rate": 8.420884176835367e-06, "loss": 0.1312, "step": 3955 }, { "epoch": 1.23, "learning_rate": 8.420484096819365e-06, "loss": 0.1091, "step": 3956 }, { "epoch": 1.23, "learning_rate": 8.420084016803362e-06, "loss": 0.1258, "step": 3957 }, { "epoch": 1.23, "learning_rate": 8.419683936787359e-06, "loss": 0.2045, "step": 3958 }, { "epoch": 1.23, "learning_rate": 8.419283856771356e-06, "loss": 0.2649, "step": 3959 }, { "epoch": 1.23, "learning_rate": 8.418883776755352e-06, "loss": 0.1707, "step": 3960 }, { "epoch": 1.23, "learning_rate": 8.418483696739349e-06, "loss": 0.2625, "step": 3961 }, { "epoch": 1.23, "learning_rate": 8.418083616723346e-06, "loss": 0.1105, "step": 3962 }, { "epoch": 1.23, "learning_rate": 8.417683536707342e-06, "loss": 0.127, "step": 3963 }, { "epoch": 1.23, "learning_rate": 8.417283456691339e-06, "loss": 0.2226, "step": 3964 }, { "epoch": 1.23, "learning_rate": 8.416883376675336e-06, "loss": 0.1433, "step": 3965 }, { "epoch": 1.23, "learning_rate": 8.416483296659332e-06, "loss": 0.2204, "step": 3966 }, { "epoch": 1.23, "learning_rate": 8.416083216643329e-06, "loss": 0.1096, "step": 3967 }, { "epoch": 1.23, "learning_rate": 8.415683136627326e-06, "loss": 0.1326, "step": 3968 }, { "epoch": 1.23, "learning_rate": 8.415283056611322e-06, "loss": 0.1966, "step": 3969 }, { "epoch": 1.23, "learning_rate": 8.41488297659532e-06, "loss": 0.149, "step": 3970 }, { "epoch": 1.23, "learning_rate": 8.414482896579317e-06, "loss": 0.2199, "step": 3971 }, { "epoch": 1.23, "learning_rate": 8.414082816563314e-06, "loss": 0.3087, "step": 3972 }, { "epoch": 1.23, "learning_rate": 8.41368273654731e-06, "loss": 0.3055, "step": 3973 }, { "epoch": 1.23, "learning_rate": 8.413282656531308e-06, "loss": 0.1545, "step": 3974 }, { "epoch": 1.23, "learning_rate": 8.412882576515303e-06, "loss": 0.2504, "step": 3975 }, { "epoch": 1.23, "learning_rate": 8.412482496499301e-06, "loss": 0.1873, "step": 3976 }, { "epoch": 1.23, "learning_rate": 8.412082416483298e-06, "loss": 0.1678, "step": 3977 }, { "epoch": 1.23, "learning_rate": 8.411682336467294e-06, "loss": 0.1, "step": 3978 }, { "epoch": 1.23, "learning_rate": 8.411282256451291e-06, "loss": 0.1757, "step": 3979 }, { "epoch": 1.24, "learning_rate": 8.410882176435288e-06, "loss": 0.1078, "step": 3980 }, { "epoch": 1.24, "learning_rate": 8.410482096419284e-06, "loss": 0.2048, "step": 3981 }, { "epoch": 1.24, "learning_rate": 8.410082016403281e-06, "loss": 0.1629, "step": 3982 }, { "epoch": 1.24, "learning_rate": 8.409681936387278e-06, "loss": 0.1377, "step": 3983 }, { "epoch": 1.24, "learning_rate": 8.409281856371276e-06, "loss": 0.0909, "step": 3984 }, { "epoch": 1.24, "learning_rate": 8.408881776355273e-06, "loss": 0.1302, "step": 3985 }, { "epoch": 1.24, "learning_rate": 8.40848169633927e-06, "loss": 0.1415, "step": 3986 }, { "epoch": 1.24, "learning_rate": 8.408081616323264e-06, "loss": 0.2018, "step": 3987 }, { "epoch": 1.24, "learning_rate": 8.407681536307261e-06, "loss": 0.2106, "step": 3988 }, { "epoch": 1.24, "learning_rate": 8.407281456291258e-06, "loss": 0.1322, "step": 3989 }, { "epoch": 1.24, "learning_rate": 8.406881376275256e-06, "loss": 0.1293, "step": 3990 }, { "epoch": 1.24, "learning_rate": 8.406481296259253e-06, "loss": 0.1026, "step": 3991 }, { "epoch": 1.24, "learning_rate": 8.40608121624325e-06, "loss": 0.1633, "step": 3992 }, { "epoch": 1.24, "learning_rate": 8.405681136227246e-06, "loss": 0.0937, "step": 3993 }, { "epoch": 1.24, "learning_rate": 8.405281056211243e-06, "loss": 0.3277, "step": 3994 }, { "epoch": 1.24, "learning_rate": 8.40488097619524e-06, "loss": 0.1604, "step": 3995 }, { "epoch": 1.24, "learning_rate": 8.404480896179236e-06, "loss": 0.168, "step": 3996 }, { "epoch": 1.24, "learning_rate": 8.404080816163233e-06, "loss": 0.0919, "step": 3997 }, { "epoch": 1.24, "learning_rate": 8.403680736147231e-06, "loss": 0.182, "step": 3998 }, { "epoch": 1.24, "learning_rate": 8.403280656131228e-06, "loss": 0.2179, "step": 3999 }, { "epoch": 1.24, "learning_rate": 8.402880576115223e-06, "loss": 0.3173, "step": 4000 }, { "epoch": 1.24, "learning_rate": 8.40248049609922e-06, "loss": 0.1506, "step": 4001 }, { "epoch": 1.24, "learning_rate": 8.402080416083216e-06, "loss": 0.088, "step": 4002 }, { "epoch": 1.24, "learning_rate": 8.401680336067213e-06, "loss": 0.2914, "step": 4003 }, { "epoch": 1.24, "learning_rate": 8.401280256051211e-06, "loss": 0.1628, "step": 4004 }, { "epoch": 1.24, "learning_rate": 8.400880176035208e-06, "loss": 0.1803, "step": 4005 }, { "epoch": 1.24, "learning_rate": 8.400480096019205e-06, "loss": 0.1844, "step": 4006 }, { "epoch": 1.24, "learning_rate": 8.400080016003202e-06, "loss": 0.1388, "step": 4007 }, { "epoch": 1.24, "learning_rate": 8.399679935987198e-06, "loss": 0.133, "step": 4008 }, { "epoch": 1.24, "learning_rate": 8.399279855971195e-06, "loss": 0.1362, "step": 4009 }, { "epoch": 1.24, "learning_rate": 8.398879775955192e-06, "loss": 0.1351, "step": 4010 }, { "epoch": 1.24, "learning_rate": 8.398479695939188e-06, "loss": 0.1883, "step": 4011 }, { "epoch": 1.25, "learning_rate": 8.398079615923185e-06, "loss": 0.2398, "step": 4012 }, { "epoch": 1.25, "learning_rate": 8.397679535907182e-06, "loss": 0.1876, "step": 4013 }, { "epoch": 1.25, "learning_rate": 8.397279455891178e-06, "loss": 0.3717, "step": 4014 }, { "epoch": 1.25, "learning_rate": 8.396879375875175e-06, "loss": 0.1378, "step": 4015 }, { "epoch": 1.25, "learning_rate": 8.396479295859172e-06, "loss": 0.1861, "step": 4016 }, { "epoch": 1.25, "learning_rate": 8.396079215843168e-06, "loss": 0.119, "step": 4017 }, { "epoch": 1.25, "learning_rate": 8.395679135827167e-06, "loss": 0.1308, "step": 4018 }, { "epoch": 1.25, "learning_rate": 8.395279055811163e-06, "loss": 0.0887, "step": 4019 }, { "epoch": 1.25, "learning_rate": 8.39487897579516e-06, "loss": 0.1434, "step": 4020 }, { "epoch": 1.25, "learning_rate": 8.394478895779157e-06, "loss": 0.1665, "step": 4021 }, { "epoch": 1.25, "learning_rate": 8.394078815763154e-06, "loss": 0.1527, "step": 4022 }, { "epoch": 1.25, "learning_rate": 8.39367873574715e-06, "loss": 0.1438, "step": 4023 }, { "epoch": 1.25, "learning_rate": 8.393278655731147e-06, "loss": 0.1431, "step": 4024 }, { "epoch": 1.25, "learning_rate": 8.392878575715144e-06, "loss": 0.2254, "step": 4025 }, { "epoch": 1.25, "learning_rate": 8.39247849569914e-06, "loss": 0.1169, "step": 4026 }, { "epoch": 1.25, "learning_rate": 8.392078415683137e-06, "loss": 0.0962, "step": 4027 }, { "epoch": 1.25, "learning_rate": 8.391678335667134e-06, "loss": 0.1945, "step": 4028 }, { "epoch": 1.25, "learning_rate": 8.39127825565113e-06, "loss": 0.0907, "step": 4029 }, { "epoch": 1.25, "learning_rate": 8.390878175635127e-06, "loss": 0.2788, "step": 4030 }, { "epoch": 1.25, "learning_rate": 8.390478095619124e-06, "loss": 0.1567, "step": 4031 }, { "epoch": 1.25, "learning_rate": 8.390078015603122e-06, "loss": 0.2302, "step": 4032 }, { "epoch": 1.25, "learning_rate": 8.389677935587119e-06, "loss": 0.1465, "step": 4033 }, { "epoch": 1.25, "learning_rate": 8.389277855571115e-06, "loss": 0.0945, "step": 4034 }, { "epoch": 1.25, "learning_rate": 8.388877775555112e-06, "loss": 0.1175, "step": 4035 }, { "epoch": 1.25, "learning_rate": 8.388477695539109e-06, "loss": 0.149, "step": 4036 }, { "epoch": 1.25, "learning_rate": 8.388077615523106e-06, "loss": 0.1568, "step": 4037 }, { "epoch": 1.25, "learning_rate": 8.387677535507102e-06, "loss": 0.1732, "step": 4038 }, { "epoch": 1.25, "learning_rate": 8.387277455491099e-06, "loss": 0.1435, "step": 4039 }, { "epoch": 1.25, "learning_rate": 8.386877375475096e-06, "loss": 0.0958, "step": 4040 }, { "epoch": 1.25, "learning_rate": 8.386477295459092e-06, "loss": 0.1669, "step": 4041 }, { "epoch": 1.25, "learning_rate": 8.386077215443089e-06, "loss": 0.1364, "step": 4042 }, { "epoch": 1.25, "learning_rate": 8.385677135427086e-06, "loss": 0.1403, "step": 4043 }, { "epoch": 1.26, "learning_rate": 8.385277055411082e-06, "loss": 0.1024, "step": 4044 }, { "epoch": 1.26, "learning_rate": 8.384876975395079e-06, "loss": 0.1963, "step": 4045 }, { "epoch": 1.26, "learning_rate": 8.384476895379077e-06, "loss": 0.1915, "step": 4046 }, { "epoch": 1.26, "learning_rate": 8.384076815363074e-06, "loss": 0.1864, "step": 4047 }, { "epoch": 1.26, "learning_rate": 8.38367673534707e-06, "loss": 0.2016, "step": 4048 }, { "epoch": 1.26, "learning_rate": 8.383276655331067e-06, "loss": 0.1835, "step": 4049 }, { "epoch": 1.26, "learning_rate": 8.382876575315062e-06, "loss": 0.18, "step": 4050 }, { "epoch": 1.26, "learning_rate": 8.38247649529906e-06, "loss": 0.1272, "step": 4051 }, { "epoch": 1.26, "learning_rate": 8.382076415283058e-06, "loss": 0.1952, "step": 4052 }, { "epoch": 1.26, "learning_rate": 8.381676335267054e-06, "loss": 0.1344, "step": 4053 }, { "epoch": 1.26, "learning_rate": 8.381276255251051e-06, "loss": 0.1525, "step": 4054 }, { "epoch": 1.26, "learning_rate": 8.380876175235048e-06, "loss": 0.29, "step": 4055 }, { "epoch": 1.26, "learning_rate": 8.380476095219044e-06, "loss": 0.1838, "step": 4056 }, { "epoch": 1.26, "learning_rate": 8.380076015203041e-06, "loss": 0.2356, "step": 4057 }, { "epoch": 1.26, "learning_rate": 8.379675935187038e-06, "loss": 0.1376, "step": 4058 }, { "epoch": 1.26, "learning_rate": 8.379275855171034e-06, "loss": 0.1558, "step": 4059 }, { "epoch": 1.26, "learning_rate": 8.378875775155033e-06, "loss": 0.1733, "step": 4060 }, { "epoch": 1.26, "learning_rate": 8.37847569513903e-06, "loss": 0.163, "step": 4061 }, { "epoch": 1.26, "learning_rate": 8.378075615123026e-06, "loss": 0.1747, "step": 4062 }, { "epoch": 1.26, "learning_rate": 8.377675535107021e-06, "loss": 0.2223, "step": 4063 }, { "epoch": 1.26, "learning_rate": 8.377275455091018e-06, "loss": 0.3132, "step": 4064 }, { "epoch": 1.26, "learning_rate": 8.376875375075016e-06, "loss": 0.0946, "step": 4065 }, { "epoch": 1.26, "learning_rate": 8.376475295059013e-06, "loss": 0.3491, "step": 4066 }, { "epoch": 1.26, "learning_rate": 8.37607521504301e-06, "loss": 0.1934, "step": 4067 }, { "epoch": 1.26, "learning_rate": 8.375675135027006e-06, "loss": 0.1531, "step": 4068 }, { "epoch": 1.26, "learning_rate": 8.375275055011003e-06, "loss": 0.1436, "step": 4069 }, { "epoch": 1.26, "learning_rate": 8.374874974995e-06, "loss": 0.1467, "step": 4070 }, { "epoch": 1.26, "learning_rate": 8.374474894978996e-06, "loss": 0.1162, "step": 4071 }, { "epoch": 1.26, "learning_rate": 8.374074814962993e-06, "loss": 0.1307, "step": 4072 }, { "epoch": 1.26, "learning_rate": 8.37367473494699e-06, "loss": 0.1778, "step": 4073 }, { "epoch": 1.26, "learning_rate": 8.373274654930988e-06, "loss": 0.2203, "step": 4074 }, { "epoch": 1.26, "learning_rate": 8.372874574914985e-06, "loss": 0.1345, "step": 4075 }, { "epoch": 1.27, "learning_rate": 8.37247449489898e-06, "loss": 0.1712, "step": 4076 }, { "epoch": 1.27, "learning_rate": 8.372074414882976e-06, "loss": 0.1681, "step": 4077 }, { "epoch": 1.27, "learning_rate": 8.371674334866973e-06, "loss": 0.225, "step": 4078 }, { "epoch": 1.27, "learning_rate": 8.371274254850971e-06, "loss": 0.1696, "step": 4079 }, { "epoch": 1.27, "learning_rate": 8.370874174834968e-06, "loss": 0.2651, "step": 4080 }, { "epoch": 1.27, "learning_rate": 8.370474094818965e-06, "loss": 0.1349, "step": 4081 }, { "epoch": 1.27, "learning_rate": 8.370074014802961e-06, "loss": 0.1005, "step": 4082 }, { "epoch": 1.27, "learning_rate": 8.369673934786958e-06, "loss": 0.3386, "step": 4083 }, { "epoch": 1.27, "learning_rate": 8.369273854770955e-06, "loss": 0.1533, "step": 4084 }, { "epoch": 1.27, "learning_rate": 8.368873774754952e-06, "loss": 0.1459, "step": 4085 }, { "epoch": 1.27, "learning_rate": 8.368473694738948e-06, "loss": 0.0405, "step": 4086 }, { "epoch": 1.27, "learning_rate": 8.368073614722945e-06, "loss": 0.1913, "step": 4087 }, { "epoch": 1.27, "learning_rate": 8.367673534706942e-06, "loss": 0.1738, "step": 4088 }, { "epoch": 1.27, "learning_rate": 8.367273454690938e-06, "loss": 0.1572, "step": 4089 }, { "epoch": 1.27, "learning_rate": 8.366873374674935e-06, "loss": 0.1188, "step": 4090 }, { "epoch": 1.27, "learning_rate": 8.366473294658932e-06, "loss": 0.2555, "step": 4091 }, { "epoch": 1.27, "learning_rate": 8.366073214642928e-06, "loss": 0.2854, "step": 4092 }, { "epoch": 1.27, "learning_rate": 8.365673134626927e-06, "loss": 0.1082, "step": 4093 }, { "epoch": 1.27, "learning_rate": 8.365273054610923e-06, "loss": 0.0578, "step": 4094 }, { "epoch": 1.27, "learning_rate": 8.36487297459492e-06, "loss": 0.1088, "step": 4095 }, { "epoch": 1.27, "learning_rate": 8.364472894578917e-06, "loss": 0.1416, "step": 4096 }, { "epoch": 1.27, "learning_rate": 8.364072814562913e-06, "loss": 0.1206, "step": 4097 }, { "epoch": 1.27, "learning_rate": 8.36367273454691e-06, "loss": 0.1203, "step": 4098 }, { "epoch": 1.27, "learning_rate": 8.363272654530907e-06, "loss": 0.0918, "step": 4099 }, { "epoch": 1.27, "learning_rate": 8.362872574514904e-06, "loss": 0.1837, "step": 4100 }, { "epoch": 1.27, "learning_rate": 8.3624724944989e-06, "loss": 0.0867, "step": 4101 }, { "epoch": 1.27, "learning_rate": 8.362072414482897e-06, "loss": 0.1082, "step": 4102 }, { "epoch": 1.27, "learning_rate": 8.361672334466894e-06, "loss": 0.0974, "step": 4103 }, { "epoch": 1.27, "learning_rate": 8.36127225445089e-06, "loss": 0.2566, "step": 4104 }, { "epoch": 1.27, "learning_rate": 8.360872174434887e-06, "loss": 0.0981, "step": 4105 }, { "epoch": 1.27, "learning_rate": 8.360472094418884e-06, "loss": 0.2554, "step": 4106 }, { "epoch": 1.27, "learning_rate": 8.360072014402882e-06, "loss": 0.1723, "step": 4107 }, { "epoch": 1.28, "learning_rate": 8.359671934386879e-06, "loss": 0.2054, "step": 4108 }, { "epoch": 1.28, "learning_rate": 8.359271854370875e-06, "loss": 0.1851, "step": 4109 }, { "epoch": 1.28, "learning_rate": 8.358871774354872e-06, "loss": 0.1701, "step": 4110 }, { "epoch": 1.28, "learning_rate": 8.358471694338869e-06, "loss": 0.1439, "step": 4111 }, { "epoch": 1.28, "learning_rate": 8.358071614322865e-06, "loss": 0.1358, "step": 4112 }, { "epoch": 1.28, "learning_rate": 8.357671534306862e-06, "loss": 0.0523, "step": 4113 }, { "epoch": 1.28, "learning_rate": 8.357271454290859e-06, "loss": 0.1096, "step": 4114 }, { "epoch": 1.28, "learning_rate": 8.356871374274856e-06, "loss": 0.139, "step": 4115 }, { "epoch": 1.28, "learning_rate": 8.356471294258852e-06, "loss": 0.2067, "step": 4116 }, { "epoch": 1.28, "learning_rate": 8.356071214242849e-06, "loss": 0.1609, "step": 4117 }, { "epoch": 1.28, "learning_rate": 8.355671134226846e-06, "loss": 0.0708, "step": 4118 }, { "epoch": 1.28, "learning_rate": 8.355271054210842e-06, "loss": 0.1018, "step": 4119 }, { "epoch": 1.28, "learning_rate": 8.354870974194839e-06, "loss": 0.1408, "step": 4120 }, { "epoch": 1.28, "learning_rate": 8.354470894178837e-06, "loss": 0.0798, "step": 4121 }, { "epoch": 1.28, "learning_rate": 8.354070814162834e-06, "loss": 0.0852, "step": 4122 }, { "epoch": 1.28, "learning_rate": 8.35367073414683e-06, "loss": 0.1728, "step": 4123 }, { "epoch": 1.28, "learning_rate": 8.353270654130827e-06, "loss": 0.194, "step": 4124 }, { "epoch": 1.28, "learning_rate": 8.352870574114824e-06, "loss": 0.1563, "step": 4125 }, { "epoch": 1.28, "learning_rate": 8.35247049409882e-06, "loss": 0.1315, "step": 4126 }, { "epoch": 1.28, "learning_rate": 8.352070414082817e-06, "loss": 0.1234, "step": 4127 }, { "epoch": 1.28, "learning_rate": 8.351670334066814e-06, "loss": 0.0827, "step": 4128 }, { "epoch": 1.28, "learning_rate": 8.35127025405081e-06, "loss": 0.2372, "step": 4129 }, { "epoch": 1.28, "learning_rate": 8.350870174034808e-06, "loss": 0.2019, "step": 4130 }, { "epoch": 1.28, "learning_rate": 8.350470094018804e-06, "loss": 0.0625, "step": 4131 }, { "epoch": 1.28, "learning_rate": 8.350070014002801e-06, "loss": 0.2419, "step": 4132 }, { "epoch": 1.28, "learning_rate": 8.349669933986798e-06, "loss": 0.1023, "step": 4133 }, { "epoch": 1.28, "learning_rate": 8.349269853970794e-06, "loss": 0.0964, "step": 4134 }, { "epoch": 1.28, "learning_rate": 8.348869773954793e-06, "loss": 0.1143, "step": 4135 }, { "epoch": 1.28, "learning_rate": 8.34846969393879e-06, "loss": 0.0783, "step": 4136 }, { "epoch": 1.28, "learning_rate": 8.348069613922786e-06, "loss": 0.2838, "step": 4137 }, { "epoch": 1.28, "learning_rate": 8.347669533906783e-06, "loss": 0.2343, "step": 4138 }, { "epoch": 1.28, "learning_rate": 8.347269453890778e-06, "loss": 0.1749, "step": 4139 }, { "epoch": 1.28, "learning_rate": 8.346869373874776e-06, "loss": 0.1394, "step": 4140 }, { "epoch": 1.29, "learning_rate": 8.346469293858773e-06, "loss": 0.1438, "step": 4141 }, { "epoch": 1.29, "learning_rate": 8.34606921384277e-06, "loss": 0.1128, "step": 4142 }, { "epoch": 1.29, "learning_rate": 8.345669133826766e-06, "loss": 0.1923, "step": 4143 }, { "epoch": 1.29, "learning_rate": 8.345269053810763e-06, "loss": 0.1203, "step": 4144 }, { "epoch": 1.29, "learning_rate": 8.34486897379476e-06, "loss": 0.281, "step": 4145 }, { "epoch": 1.29, "learning_rate": 8.344468893778756e-06, "loss": 0.2182, "step": 4146 }, { "epoch": 1.29, "learning_rate": 8.344068813762753e-06, "loss": 0.1941, "step": 4147 }, { "epoch": 1.29, "learning_rate": 8.34366873374675e-06, "loss": 0.2681, "step": 4148 }, { "epoch": 1.29, "learning_rate": 8.343268653730748e-06, "loss": 0.1252, "step": 4149 }, { "epoch": 1.29, "learning_rate": 8.342868573714745e-06, "loss": 0.1749, "step": 4150 }, { "epoch": 1.29, "learning_rate": 8.34246849369874e-06, "loss": 0.1475, "step": 4151 }, { "epoch": 1.29, "learning_rate": 8.342068413682736e-06, "loss": 0.204, "step": 4152 }, { "epoch": 1.29, "learning_rate": 8.341668333666733e-06, "loss": 0.1812, "step": 4153 }, { "epoch": 1.29, "learning_rate": 8.341268253650731e-06, "loss": 0.1233, "step": 4154 }, { "epoch": 1.29, "learning_rate": 8.340868173634728e-06, "loss": 0.1202, "step": 4155 }, { "epoch": 1.29, "learning_rate": 8.340468093618725e-06, "loss": 0.0527, "step": 4156 }, { "epoch": 1.29, "learning_rate": 8.340068013602721e-06, "loss": 0.2188, "step": 4157 }, { "epoch": 1.29, "learning_rate": 8.339667933586718e-06, "loss": 0.163, "step": 4158 }, { "epoch": 1.29, "learning_rate": 8.339267853570715e-06, "loss": 0.1577, "step": 4159 }, { "epoch": 1.29, "learning_rate": 8.338867773554711e-06, "loss": 0.1769, "step": 4160 }, { "epoch": 1.29, "learning_rate": 8.338467693538708e-06, "loss": 0.06, "step": 4161 }, { "epoch": 1.29, "learning_rate": 8.338067613522705e-06, "loss": 0.1247, "step": 4162 }, { "epoch": 1.29, "learning_rate": 8.337667533506703e-06, "loss": 0.1098, "step": 4163 }, { "epoch": 1.29, "learning_rate": 8.337267453490698e-06, "loss": 0.2979, "step": 4164 }, { "epoch": 1.29, "learning_rate": 8.336867373474695e-06, "loss": 0.1502, "step": 4165 }, { "epoch": 1.29, "learning_rate": 8.336467293458692e-06, "loss": 0.1066, "step": 4166 }, { "epoch": 1.29, "learning_rate": 8.336067213442688e-06, "loss": 0.2028, "step": 4167 }, { "epoch": 1.29, "learning_rate": 8.335667133426687e-06, "loss": 0.2174, "step": 4168 }, { "epoch": 1.29, "learning_rate": 8.335267053410683e-06, "loss": 0.1615, "step": 4169 }, { "epoch": 1.29, "learning_rate": 8.33486697339468e-06, "loss": 0.1239, "step": 4170 }, { "epoch": 1.29, "learning_rate": 8.334466893378677e-06, "loss": 0.2149, "step": 4171 }, { "epoch": 1.29, "learning_rate": 8.334066813362673e-06, "loss": 0.1932, "step": 4172 }, { "epoch": 1.3, "learning_rate": 8.33366673334667e-06, "loss": 0.1775, "step": 4173 }, { "epoch": 1.3, "learning_rate": 8.333266653330667e-06, "loss": 0.1619, "step": 4174 }, { "epoch": 1.3, "learning_rate": 8.332866573314663e-06, "loss": 0.0878, "step": 4175 }, { "epoch": 1.3, "learning_rate": 8.33246649329866e-06, "loss": 0.0869, "step": 4176 }, { "epoch": 1.3, "learning_rate": 8.332066413282657e-06, "loss": 0.1457, "step": 4177 }, { "epoch": 1.3, "learning_rate": 8.331666333266654e-06, "loss": 0.2385, "step": 4178 }, { "epoch": 1.3, "learning_rate": 8.33126625325065e-06, "loss": 0.102, "step": 4179 }, { "epoch": 1.3, "learning_rate": 8.330866173234647e-06, "loss": 0.1527, "step": 4180 }, { "epoch": 1.3, "learning_rate": 8.330466093218644e-06, "loss": 0.1934, "step": 4181 }, { "epoch": 1.3, "learning_rate": 8.330066013202642e-06, "loss": 0.2248, "step": 4182 }, { "epoch": 1.3, "learning_rate": 8.329665933186639e-06, "loss": 0.1468, "step": 4183 }, { "epoch": 1.3, "learning_rate": 8.329265853170635e-06, "loss": 0.1682, "step": 4184 }, { "epoch": 1.3, "learning_rate": 8.328865773154632e-06, "loss": 0.1907, "step": 4185 }, { "epoch": 1.3, "learning_rate": 8.328465693138629e-06, "loss": 0.1674, "step": 4186 }, { "epoch": 1.3, "learning_rate": 8.328065613122625e-06, "loss": 0.2219, "step": 4187 }, { "epoch": 1.3, "learning_rate": 8.327665533106622e-06, "loss": 0.1184, "step": 4188 }, { "epoch": 1.3, "learning_rate": 8.327265453090619e-06, "loss": 0.2277, "step": 4189 }, { "epoch": 1.3, "learning_rate": 8.326865373074615e-06, "loss": 0.1513, "step": 4190 }, { "epoch": 1.3, "learning_rate": 8.326465293058612e-06, "loss": 0.1769, "step": 4191 }, { "epoch": 1.3, "learning_rate": 8.326065213042609e-06, "loss": 0.1813, "step": 4192 }, { "epoch": 1.3, "learning_rate": 8.325665133026606e-06, "loss": 0.1036, "step": 4193 }, { "epoch": 1.3, "learning_rate": 8.325265053010602e-06, "loss": 0.0965, "step": 4194 }, { "epoch": 1.3, "learning_rate": 8.324864972994599e-06, "loss": 0.267, "step": 4195 }, { "epoch": 1.3, "learning_rate": 8.324464892978597e-06, "loss": 0.2534, "step": 4196 }, { "epoch": 1.3, "learning_rate": 8.324064812962594e-06, "loss": 0.1764, "step": 4197 }, { "epoch": 1.3, "learning_rate": 8.32366473294659e-06, "loss": 0.2378, "step": 4198 }, { "epoch": 1.3, "learning_rate": 8.323264652930587e-06, "loss": 0.178, "step": 4199 }, { "epoch": 1.3, "learning_rate": 8.322864572914584e-06, "loss": 0.1134, "step": 4200 }, { "epoch": 1.3, "learning_rate": 8.32246449289858e-06, "loss": 0.1922, "step": 4201 }, { "epoch": 1.3, "learning_rate": 8.322064412882577e-06, "loss": 0.175, "step": 4202 }, { "epoch": 1.3, "learning_rate": 8.321664332866574e-06, "loss": 0.1887, "step": 4203 }, { "epoch": 1.3, "learning_rate": 8.32126425285057e-06, "loss": 0.1233, "step": 4204 }, { "epoch": 1.31, "learning_rate": 8.320864172834567e-06, "loss": 0.2403, "step": 4205 }, { "epoch": 1.31, "learning_rate": 8.320464092818564e-06, "loss": 0.1551, "step": 4206 }, { "epoch": 1.31, "learning_rate": 8.32006401280256e-06, "loss": 0.1931, "step": 4207 }, { "epoch": 1.31, "learning_rate": 8.319663932786558e-06, "loss": 0.1121, "step": 4208 }, { "epoch": 1.31, "learning_rate": 8.319263852770554e-06, "loss": 0.1418, "step": 4209 }, { "epoch": 1.31, "learning_rate": 8.318863772754553e-06, "loss": 0.1224, "step": 4210 }, { "epoch": 1.31, "learning_rate": 8.31846369273855e-06, "loss": 0.1481, "step": 4211 }, { "epoch": 1.31, "learning_rate": 8.318063612722546e-06, "loss": 0.1367, "step": 4212 }, { "epoch": 1.31, "learning_rate": 8.317663532706543e-06, "loss": 0.1429, "step": 4213 }, { "epoch": 1.31, "learning_rate": 8.317263452690538e-06, "loss": 0.1769, "step": 4214 }, { "epoch": 1.31, "learning_rate": 8.316863372674534e-06, "loss": 0.1414, "step": 4215 }, { "epoch": 1.31, "learning_rate": 8.316463292658533e-06, "loss": 0.1362, "step": 4216 }, { "epoch": 1.31, "learning_rate": 8.31606321264253e-06, "loss": 0.0927, "step": 4217 }, { "epoch": 1.31, "learning_rate": 8.315663132626526e-06, "loss": 0.1029, "step": 4218 }, { "epoch": 1.31, "learning_rate": 8.315263052610523e-06, "loss": 0.1687, "step": 4219 }, { "epoch": 1.31, "learning_rate": 8.31486297259452e-06, "loss": 0.1437, "step": 4220 }, { "epoch": 1.31, "learning_rate": 8.314462892578516e-06, "loss": 0.1704, "step": 4221 }, { "epoch": 1.31, "learning_rate": 8.314062812562513e-06, "loss": 0.1957, "step": 4222 }, { "epoch": 1.31, "learning_rate": 8.31366273254651e-06, "loss": 0.1276, "step": 4223 }, { "epoch": 1.31, "learning_rate": 8.313262652530508e-06, "loss": 0.168, "step": 4224 }, { "epoch": 1.31, "learning_rate": 8.312862572514505e-06, "loss": 0.1816, "step": 4225 }, { "epoch": 1.31, "learning_rate": 8.312462492498501e-06, "loss": 0.2321, "step": 4226 }, { "epoch": 1.31, "learning_rate": 8.312062412482496e-06, "loss": 0.1262, "step": 4227 }, { "epoch": 1.31, "learning_rate": 8.311662332466493e-06, "loss": 0.1908, "step": 4228 }, { "epoch": 1.31, "learning_rate": 8.31126225245049e-06, "loss": 0.1949, "step": 4229 }, { "epoch": 1.31, "learning_rate": 8.310862172434488e-06, "loss": 0.0899, "step": 4230 }, { "epoch": 1.31, "learning_rate": 8.310462092418485e-06, "loss": 0.0921, "step": 4231 }, { "epoch": 1.31, "learning_rate": 8.310062012402481e-06, "loss": 0.1665, "step": 4232 }, { "epoch": 1.31, "learning_rate": 8.309661932386478e-06, "loss": 0.1395, "step": 4233 }, { "epoch": 1.31, "learning_rate": 8.309261852370475e-06, "loss": 0.0513, "step": 4234 }, { "epoch": 1.31, "learning_rate": 8.308861772354471e-06, "loss": 0.1532, "step": 4235 }, { "epoch": 1.31, "learning_rate": 8.308461692338468e-06, "loss": 0.1511, "step": 4236 }, { "epoch": 1.32, "learning_rate": 8.308061612322465e-06, "loss": 0.1088, "step": 4237 }, { "epoch": 1.32, "learning_rate": 8.307661532306463e-06, "loss": 0.2209, "step": 4238 }, { "epoch": 1.32, "learning_rate": 8.307261452290458e-06, "loss": 0.1836, "step": 4239 }, { "epoch": 1.32, "learning_rate": 8.306861372274455e-06, "loss": 0.2097, "step": 4240 }, { "epoch": 1.32, "learning_rate": 8.306461292258452e-06, "loss": 0.172, "step": 4241 }, { "epoch": 1.32, "learning_rate": 8.306061212242448e-06, "loss": 0.1155, "step": 4242 }, { "epoch": 1.32, "learning_rate": 8.305661132226445e-06, "loss": 0.1491, "step": 4243 }, { "epoch": 1.32, "learning_rate": 8.305261052210443e-06, "loss": 0.1512, "step": 4244 }, { "epoch": 1.32, "learning_rate": 8.30486097219444e-06, "loss": 0.2657, "step": 4245 }, { "epoch": 1.32, "learning_rate": 8.304460892178437e-06, "loss": 0.105, "step": 4246 }, { "epoch": 1.32, "learning_rate": 8.304060812162433e-06, "loss": 0.1493, "step": 4247 }, { "epoch": 1.32, "learning_rate": 8.30366073214643e-06, "loss": 0.1603, "step": 4248 }, { "epoch": 1.32, "learning_rate": 8.303260652130427e-06, "loss": 0.1724, "step": 4249 }, { "epoch": 1.32, "learning_rate": 8.302860572114423e-06, "loss": 0.11, "step": 4250 }, { "epoch": 1.32, "learning_rate": 8.30246049209842e-06, "loss": 0.1005, "step": 4251 }, { "epoch": 1.32, "learning_rate": 8.302060412082417e-06, "loss": 0.1623, "step": 4252 }, { "epoch": 1.32, "learning_rate": 8.301660332066413e-06, "loss": 0.1043, "step": 4253 }, { "epoch": 1.32, "learning_rate": 8.30126025205041e-06, "loss": 0.0843, "step": 4254 }, { "epoch": 1.32, "learning_rate": 8.300860172034407e-06, "loss": 0.083, "step": 4255 }, { "epoch": 1.32, "learning_rate": 8.300460092018404e-06, "loss": 0.1404, "step": 4256 }, { "epoch": 1.32, "learning_rate": 8.300060012002402e-06, "loss": 0.1696, "step": 4257 }, { "epoch": 1.32, "learning_rate": 8.299659931986399e-06, "loss": 0.1729, "step": 4258 }, { "epoch": 1.32, "learning_rate": 8.299259851970395e-06, "loss": 0.0958, "step": 4259 }, { "epoch": 1.32, "learning_rate": 8.298859771954392e-06, "loss": 0.2194, "step": 4260 }, { "epoch": 1.32, "learning_rate": 8.298459691938389e-06, "loss": 0.2233, "step": 4261 }, { "epoch": 1.32, "learning_rate": 8.298059611922385e-06, "loss": 0.1366, "step": 4262 }, { "epoch": 1.32, "learning_rate": 8.297659531906382e-06, "loss": 0.1334, "step": 4263 }, { "epoch": 1.32, "learning_rate": 8.297259451890379e-06, "loss": 0.094, "step": 4264 }, { "epoch": 1.32, "learning_rate": 8.296859371874375e-06, "loss": 0.1621, "step": 4265 }, { "epoch": 1.32, "learning_rate": 8.296459291858372e-06, "loss": 0.1354, "step": 4266 }, { "epoch": 1.32, "learning_rate": 8.296059211842369e-06, "loss": 0.2337, "step": 4267 }, { "epoch": 1.32, "learning_rate": 8.295659131826365e-06, "loss": 0.1234, "step": 4268 }, { "epoch": 1.33, "learning_rate": 8.295259051810362e-06, "loss": 0.1614, "step": 4269 }, { "epoch": 1.33, "learning_rate": 8.294858971794359e-06, "loss": 0.1568, "step": 4270 }, { "epoch": 1.33, "learning_rate": 8.294458891778357e-06, "loss": 0.0758, "step": 4271 }, { "epoch": 1.33, "learning_rate": 8.294058811762354e-06, "loss": 0.1872, "step": 4272 }, { "epoch": 1.33, "learning_rate": 8.29365873174635e-06, "loss": 0.1326, "step": 4273 }, { "epoch": 1.33, "learning_rate": 8.293258651730347e-06, "loss": 0.1392, "step": 4274 }, { "epoch": 1.33, "learning_rate": 8.292858571714344e-06, "loss": 0.0664, "step": 4275 }, { "epoch": 1.33, "learning_rate": 8.29245849169834e-06, "loss": 0.0805, "step": 4276 }, { "epoch": 1.33, "learning_rate": 8.292058411682337e-06, "loss": 0.1042, "step": 4277 }, { "epoch": 1.33, "learning_rate": 8.291658331666334e-06, "loss": 0.1713, "step": 4278 }, { "epoch": 1.33, "learning_rate": 8.29125825165033e-06, "loss": 0.1151, "step": 4279 }, { "epoch": 1.33, "learning_rate": 8.290858171634327e-06, "loss": 0.1106, "step": 4280 }, { "epoch": 1.33, "learning_rate": 8.290458091618324e-06, "loss": 0.1569, "step": 4281 }, { "epoch": 1.33, "learning_rate": 8.29005801160232e-06, "loss": 0.2081, "step": 4282 }, { "epoch": 1.33, "learning_rate": 8.289657931586317e-06, "loss": 0.1822, "step": 4283 }, { "epoch": 1.33, "learning_rate": 8.289257851570314e-06, "loss": 0.1446, "step": 4284 }, { "epoch": 1.33, "learning_rate": 8.288857771554313e-06, "loss": 0.2012, "step": 4285 }, { "epoch": 1.33, "learning_rate": 8.28845769153831e-06, "loss": 0.0494, "step": 4286 }, { "epoch": 1.33, "learning_rate": 8.288057611522306e-06, "loss": 0.2012, "step": 4287 }, { "epoch": 1.33, "learning_rate": 8.287657531506303e-06, "loss": 0.1202, "step": 4288 }, { "epoch": 1.33, "learning_rate": 8.2872574514903e-06, "loss": 0.2295, "step": 4289 }, { "epoch": 1.33, "learning_rate": 8.286857371474294e-06, "loss": 0.1172, "step": 4290 }, { "epoch": 1.33, "learning_rate": 8.286457291458293e-06, "loss": 0.2126, "step": 4291 }, { "epoch": 1.33, "learning_rate": 8.28605721144229e-06, "loss": 0.1425, "step": 4292 }, { "epoch": 1.33, "learning_rate": 8.285657131426286e-06, "loss": 0.2181, "step": 4293 }, { "epoch": 1.33, "learning_rate": 8.285257051410283e-06, "loss": 0.2248, "step": 4294 }, { "epoch": 1.33, "learning_rate": 8.28485697139428e-06, "loss": 0.3171, "step": 4295 }, { "epoch": 1.33, "learning_rate": 8.284456891378276e-06, "loss": 0.1255, "step": 4296 }, { "epoch": 1.33, "learning_rate": 8.284056811362273e-06, "loss": 0.2088, "step": 4297 }, { "epoch": 1.33, "learning_rate": 8.28365673134627e-06, "loss": 0.0662, "step": 4298 }, { "epoch": 1.33, "learning_rate": 8.283256651330268e-06, "loss": 0.2269, "step": 4299 }, { "epoch": 1.33, "learning_rate": 8.282856571314265e-06, "loss": 0.1959, "step": 4300 }, { "epoch": 1.33, "learning_rate": 8.282456491298261e-06, "loss": 0.1047, "step": 4301 }, { "epoch": 1.34, "learning_rate": 8.282056411282258e-06, "loss": 0.1403, "step": 4302 }, { "epoch": 1.34, "learning_rate": 8.281656331266253e-06, "loss": 0.1912, "step": 4303 }, { "epoch": 1.34, "learning_rate": 8.28125625125025e-06, "loss": 0.1229, "step": 4304 }, { "epoch": 1.34, "learning_rate": 8.280856171234248e-06, "loss": 0.143, "step": 4305 }, { "epoch": 1.34, "learning_rate": 8.280456091218245e-06, "loss": 0.1783, "step": 4306 }, { "epoch": 1.34, "learning_rate": 8.280056011202241e-06, "loss": 0.1668, "step": 4307 }, { "epoch": 1.34, "learning_rate": 8.279655931186238e-06, "loss": 0.064, "step": 4308 }, { "epoch": 1.34, "learning_rate": 8.279255851170235e-06, "loss": 0.0855, "step": 4309 }, { "epoch": 1.34, "learning_rate": 8.278855771154231e-06, "loss": 0.2517, "step": 4310 }, { "epoch": 1.34, "learning_rate": 8.278455691138228e-06, "loss": 0.2656, "step": 4311 }, { "epoch": 1.34, "learning_rate": 8.278055611122225e-06, "loss": 0.103, "step": 4312 }, { "epoch": 1.34, "learning_rate": 8.277655531106223e-06, "loss": 0.1316, "step": 4313 }, { "epoch": 1.34, "learning_rate": 8.27725545109022e-06, "loss": 0.136, "step": 4314 }, { "epoch": 1.34, "learning_rate": 8.276855371074215e-06, "loss": 0.2103, "step": 4315 }, { "epoch": 1.34, "learning_rate": 8.276455291058211e-06, "loss": 0.1237, "step": 4316 }, { "epoch": 1.34, "learning_rate": 8.276055211042208e-06, "loss": 0.1513, "step": 4317 }, { "epoch": 1.34, "learning_rate": 8.275655131026205e-06, "loss": 0.0693, "step": 4318 }, { "epoch": 1.34, "learning_rate": 8.275255051010203e-06, "loss": 0.2348, "step": 4319 }, { "epoch": 1.34, "learning_rate": 8.2748549709942e-06, "loss": 0.0948, "step": 4320 }, { "epoch": 1.34, "learning_rate": 8.274454890978197e-06, "loss": 0.2304, "step": 4321 }, { "epoch": 1.34, "learning_rate": 8.274054810962193e-06, "loss": 0.1538, "step": 4322 }, { "epoch": 1.34, "learning_rate": 8.27365473094619e-06, "loss": 0.1173, "step": 4323 }, { "epoch": 1.34, "learning_rate": 8.273254650930187e-06, "loss": 0.1323, "step": 4324 }, { "epoch": 1.34, "learning_rate": 8.272854570914183e-06, "loss": 0.1612, "step": 4325 }, { "epoch": 1.34, "learning_rate": 8.27245449089818e-06, "loss": 0.1806, "step": 4326 }, { "epoch": 1.34, "learning_rate": 8.272054410882178e-06, "loss": 0.1447, "step": 4327 }, { "epoch": 1.34, "learning_rate": 8.271654330866173e-06, "loss": 0.1011, "step": 4328 }, { "epoch": 1.34, "learning_rate": 8.27125425085017e-06, "loss": 0.0911, "step": 4329 }, { "epoch": 1.34, "learning_rate": 8.270854170834167e-06, "loss": 0.1829, "step": 4330 }, { "epoch": 1.34, "learning_rate": 8.270454090818163e-06, "loss": 0.125, "step": 4331 }, { "epoch": 1.34, "learning_rate": 8.27005401080216e-06, "loss": 0.1927, "step": 4332 }, { "epoch": 1.34, "learning_rate": 8.269653930786159e-06, "loss": 0.1244, "step": 4333 }, { "epoch": 1.35, "learning_rate": 8.269253850770155e-06, "loss": 0.1664, "step": 4334 }, { "epoch": 1.35, "learning_rate": 8.268853770754152e-06, "loss": 0.1819, "step": 4335 }, { "epoch": 1.35, "learning_rate": 8.268453690738149e-06, "loss": 0.2056, "step": 4336 }, { "epoch": 1.35, "learning_rate": 8.268053610722145e-06, "loss": 0.1852, "step": 4337 }, { "epoch": 1.35, "learning_rate": 8.267653530706142e-06, "loss": 0.1508, "step": 4338 }, { "epoch": 1.35, "learning_rate": 8.267253450690139e-06, "loss": 0.1307, "step": 4339 }, { "epoch": 1.35, "learning_rate": 8.266853370674135e-06, "loss": 0.0628, "step": 4340 }, { "epoch": 1.35, "learning_rate": 8.266453290658132e-06, "loss": 0.106, "step": 4341 }, { "epoch": 1.35, "learning_rate": 8.266053210642129e-06, "loss": 0.1333, "step": 4342 }, { "epoch": 1.35, "learning_rate": 8.265653130626125e-06, "loss": 0.1949, "step": 4343 }, { "epoch": 1.35, "learning_rate": 8.265253050610122e-06, "loss": 0.1231, "step": 4344 }, { "epoch": 1.35, "learning_rate": 8.264852970594119e-06, "loss": 0.1317, "step": 4345 }, { "epoch": 1.35, "learning_rate": 8.264452890578115e-06, "loss": 0.208, "step": 4346 }, { "epoch": 1.35, "learning_rate": 8.264052810562114e-06, "loss": 0.1414, "step": 4347 }, { "epoch": 1.35, "learning_rate": 8.26365273054611e-06, "loss": 0.1263, "step": 4348 }, { "epoch": 1.35, "learning_rate": 8.263252650530107e-06, "loss": 0.4192, "step": 4349 }, { "epoch": 1.35, "learning_rate": 8.262852570514104e-06, "loss": 0.146, "step": 4350 }, { "epoch": 1.35, "learning_rate": 8.2624524904981e-06, "loss": 0.2887, "step": 4351 }, { "epoch": 1.35, "learning_rate": 8.262052410482097e-06, "loss": 0.1797, "step": 4352 }, { "epoch": 1.35, "learning_rate": 8.261652330466094e-06, "loss": 0.1878, "step": 4353 }, { "epoch": 1.35, "learning_rate": 8.26125225045009e-06, "loss": 0.1264, "step": 4354 }, { "epoch": 1.35, "learning_rate": 8.260852170434087e-06, "loss": 0.1489, "step": 4355 }, { "epoch": 1.35, "learning_rate": 8.260452090418084e-06, "loss": 0.3121, "step": 4356 }, { "epoch": 1.35, "learning_rate": 8.26005201040208e-06, "loss": 0.2434, "step": 4357 }, { "epoch": 1.35, "learning_rate": 8.259651930386077e-06, "loss": 0.1943, "step": 4358 }, { "epoch": 1.35, "learning_rate": 8.259251850370074e-06, "loss": 0.0696, "step": 4359 }, { "epoch": 1.35, "learning_rate": 8.25885177035407e-06, "loss": 0.0934, "step": 4360 }, { "epoch": 1.35, "learning_rate": 8.258451690338069e-06, "loss": 0.1928, "step": 4361 }, { "epoch": 1.35, "learning_rate": 8.258051610322066e-06, "loss": 0.3186, "step": 4362 }, { "epoch": 1.35, "learning_rate": 8.257651530306063e-06, "loss": 0.131, "step": 4363 }, { "epoch": 1.35, "learning_rate": 8.25725145029006e-06, "loss": 0.1162, "step": 4364 }, { "epoch": 1.35, "learning_rate": 8.256851370274056e-06, "loss": 0.1125, "step": 4365 }, { "epoch": 1.36, "learning_rate": 8.256451290258053e-06, "loss": 0.1425, "step": 4366 }, { "epoch": 1.36, "learning_rate": 8.25605121024205e-06, "loss": 0.1442, "step": 4367 }, { "epoch": 1.36, "learning_rate": 8.255651130226046e-06, "loss": 0.1704, "step": 4368 }, { "epoch": 1.36, "learning_rate": 8.255251050210043e-06, "loss": 0.2061, "step": 4369 }, { "epoch": 1.36, "learning_rate": 8.25485097019404e-06, "loss": 0.2128, "step": 4370 }, { "epoch": 1.36, "learning_rate": 8.254450890178036e-06, "loss": 0.1962, "step": 4371 }, { "epoch": 1.36, "learning_rate": 8.254050810162033e-06, "loss": 0.107, "step": 4372 }, { "epoch": 1.36, "learning_rate": 8.25365073014603e-06, "loss": 0.1867, "step": 4373 }, { "epoch": 1.36, "learning_rate": 8.253250650130026e-06, "loss": 0.1502, "step": 4374 }, { "epoch": 1.36, "learning_rate": 8.252850570114024e-06, "loss": 0.1614, "step": 4375 }, { "epoch": 1.36, "learning_rate": 8.252450490098021e-06, "loss": 0.132, "step": 4376 }, { "epoch": 1.36, "learning_rate": 8.252050410082018e-06, "loss": 0.1356, "step": 4377 }, { "epoch": 1.36, "learning_rate": 8.251650330066013e-06, "loss": 0.1733, "step": 4378 }, { "epoch": 1.36, "learning_rate": 8.25125025005001e-06, "loss": 0.3164, "step": 4379 }, { "epoch": 1.36, "learning_rate": 8.250850170034008e-06, "loss": 0.2164, "step": 4380 }, { "epoch": 1.36, "learning_rate": 8.250450090018005e-06, "loss": 0.2407, "step": 4381 }, { "epoch": 1.36, "learning_rate": 8.250050010002001e-06, "loss": 0.2775, "step": 4382 }, { "epoch": 1.36, "learning_rate": 8.249649929985998e-06, "loss": 0.1204, "step": 4383 }, { "epoch": 1.36, "learning_rate": 8.249249849969995e-06, "loss": 0.1879, "step": 4384 }, { "epoch": 1.36, "learning_rate": 8.248849769953991e-06, "loss": 0.2284, "step": 4385 }, { "epoch": 1.36, "learning_rate": 8.248449689937988e-06, "loss": 0.1471, "step": 4386 }, { "epoch": 1.36, "learning_rate": 8.248049609921985e-06, "loss": 0.1951, "step": 4387 }, { "epoch": 1.36, "learning_rate": 8.247649529905981e-06, "loss": 0.1653, "step": 4388 }, { "epoch": 1.36, "learning_rate": 8.24724944988998e-06, "loss": 0.092, "step": 4389 }, { "epoch": 1.36, "learning_rate": 8.246849369873976e-06, "loss": 0.1921, "step": 4390 }, { "epoch": 1.36, "learning_rate": 8.246449289857971e-06, "loss": 0.1362, "step": 4391 }, { "epoch": 1.36, "learning_rate": 8.246049209841968e-06, "loss": 0.1526, "step": 4392 }, { "epoch": 1.36, "learning_rate": 8.245649129825965e-06, "loss": 0.1252, "step": 4393 }, { "epoch": 1.36, "learning_rate": 8.245249049809963e-06, "loss": 0.1996, "step": 4394 }, { "epoch": 1.36, "learning_rate": 8.24484896979396e-06, "loss": 0.0774, "step": 4395 }, { "epoch": 1.36, "learning_rate": 8.244448889777957e-06, "loss": 0.1691, "step": 4396 }, { "epoch": 1.36, "learning_rate": 8.244048809761953e-06, "loss": 0.1991, "step": 4397 }, { "epoch": 1.37, "learning_rate": 8.24364872974595e-06, "loss": 0.2281, "step": 4398 }, { "epoch": 1.37, "learning_rate": 8.243248649729947e-06, "loss": 0.1223, "step": 4399 }, { "epoch": 1.37, "learning_rate": 8.242848569713943e-06, "loss": 0.201, "step": 4400 }, { "epoch": 1.37, "learning_rate": 8.24244848969794e-06, "loss": 0.0929, "step": 4401 }, { "epoch": 1.37, "learning_rate": 8.242048409681938e-06, "loss": 0.1447, "step": 4402 }, { "epoch": 1.37, "learning_rate": 8.241648329665933e-06, "loss": 0.1479, "step": 4403 }, { "epoch": 1.37, "learning_rate": 8.24124824964993e-06, "loss": 0.3029, "step": 4404 }, { "epoch": 1.37, "learning_rate": 8.240848169633927e-06, "loss": 0.1534, "step": 4405 }, { "epoch": 1.37, "learning_rate": 8.240448089617923e-06, "loss": 0.1106, "step": 4406 }, { "epoch": 1.37, "learning_rate": 8.24004800960192e-06, "loss": 0.243, "step": 4407 }, { "epoch": 1.37, "learning_rate": 8.239647929585918e-06, "loss": 0.1058, "step": 4408 }, { "epoch": 1.37, "learning_rate": 8.239247849569915e-06, "loss": 0.1093, "step": 4409 }, { "epoch": 1.37, "learning_rate": 8.238847769553912e-06, "loss": 0.1881, "step": 4410 }, { "epoch": 1.37, "learning_rate": 8.238447689537909e-06, "loss": 0.23, "step": 4411 }, { "epoch": 1.37, "learning_rate": 8.238047609521905e-06, "loss": 0.2458, "step": 4412 }, { "epoch": 1.37, "learning_rate": 8.237647529505902e-06, "loss": 0.1275, "step": 4413 }, { "epoch": 1.37, "learning_rate": 8.237247449489899e-06, "loss": 0.0992, "step": 4414 }, { "epoch": 1.37, "learning_rate": 8.236847369473895e-06, "loss": 0.165, "step": 4415 }, { "epoch": 1.37, "learning_rate": 8.236447289457892e-06, "loss": 0.1184, "step": 4416 }, { "epoch": 1.37, "learning_rate": 8.236047209441889e-06, "loss": 0.1138, "step": 4417 }, { "epoch": 1.37, "learning_rate": 8.235647129425885e-06, "loss": 0.1334, "step": 4418 }, { "epoch": 1.37, "learning_rate": 8.235247049409882e-06, "loss": 0.1769, "step": 4419 }, { "epoch": 1.37, "learning_rate": 8.234846969393879e-06, "loss": 0.1667, "step": 4420 }, { "epoch": 1.37, "learning_rate": 8.234446889377875e-06, "loss": 0.1913, "step": 4421 }, { "epoch": 1.37, "learning_rate": 8.234046809361874e-06, "loss": 0.1907, "step": 4422 }, { "epoch": 1.37, "learning_rate": 8.23364672934587e-06, "loss": 0.1641, "step": 4423 }, { "epoch": 1.37, "learning_rate": 8.233246649329867e-06, "loss": 0.2092, "step": 4424 }, { "epoch": 1.37, "learning_rate": 8.232846569313864e-06, "loss": 0.172, "step": 4425 }, { "epoch": 1.37, "learning_rate": 8.23244648929786e-06, "loss": 0.1744, "step": 4426 }, { "epoch": 1.37, "learning_rate": 8.232046409281857e-06, "loss": 0.2712, "step": 4427 }, { "epoch": 1.37, "learning_rate": 8.231646329265854e-06, "loss": 0.1838, "step": 4428 }, { "epoch": 1.37, "learning_rate": 8.23124624924985e-06, "loss": 0.1673, "step": 4429 }, { "epoch": 1.37, "learning_rate": 8.230846169233847e-06, "loss": 0.1078, "step": 4430 }, { "epoch": 1.38, "learning_rate": 8.230446089217844e-06, "loss": 0.1195, "step": 4431 }, { "epoch": 1.38, "learning_rate": 8.23004600920184e-06, "loss": 0.1241, "step": 4432 }, { "epoch": 1.38, "learning_rate": 8.229645929185837e-06, "loss": 0.2107, "step": 4433 }, { "epoch": 1.38, "learning_rate": 8.229245849169834e-06, "loss": 0.0977, "step": 4434 }, { "epoch": 1.38, "learning_rate": 8.22884576915383e-06, "loss": 0.1944, "step": 4435 }, { "epoch": 1.38, "learning_rate": 8.228445689137829e-06, "loss": 0.1878, "step": 4436 }, { "epoch": 1.38, "learning_rate": 8.228045609121826e-06, "loss": 0.2228, "step": 4437 }, { "epoch": 1.38, "learning_rate": 8.227645529105822e-06, "loss": 0.1491, "step": 4438 }, { "epoch": 1.38, "learning_rate": 8.227245449089819e-06, "loss": 0.065, "step": 4439 }, { "epoch": 1.38, "learning_rate": 8.226845369073816e-06, "loss": 0.1485, "step": 4440 }, { "epoch": 1.38, "learning_rate": 8.226445289057813e-06, "loss": 0.202, "step": 4441 }, { "epoch": 1.38, "learning_rate": 8.22604520904181e-06, "loss": 0.1985, "step": 4442 }, { "epoch": 1.38, "learning_rate": 8.225645129025806e-06, "loss": 0.3226, "step": 4443 }, { "epoch": 1.38, "learning_rate": 8.225245049009803e-06, "loss": 0.1298, "step": 4444 }, { "epoch": 1.38, "learning_rate": 8.2248449689938e-06, "loss": 0.1032, "step": 4445 }, { "epoch": 1.38, "learning_rate": 8.224444888977796e-06, "loss": 0.0951, "step": 4446 }, { "epoch": 1.38, "learning_rate": 8.224044808961793e-06, "loss": 0.1036, "step": 4447 }, { "epoch": 1.38, "learning_rate": 8.22364472894579e-06, "loss": 0.201, "step": 4448 }, { "epoch": 1.38, "learning_rate": 8.223244648929786e-06, "loss": 0.122, "step": 4449 }, { "epoch": 1.38, "learning_rate": 8.222844568913784e-06, "loss": 0.2975, "step": 4450 }, { "epoch": 1.38, "learning_rate": 8.222444488897781e-06, "loss": 0.1348, "step": 4451 }, { "epoch": 1.38, "learning_rate": 8.222044408881778e-06, "loss": 0.1469, "step": 4452 }, { "epoch": 1.38, "learning_rate": 8.221644328865774e-06, "loss": 0.1182, "step": 4453 }, { "epoch": 1.38, "learning_rate": 8.22124424884977e-06, "loss": 0.136, "step": 4454 }, { "epoch": 1.38, "learning_rate": 8.220844168833768e-06, "loss": 0.1309, "step": 4455 }, { "epoch": 1.38, "learning_rate": 8.220444088817765e-06, "loss": 0.1081, "step": 4456 }, { "epoch": 1.38, "learning_rate": 8.220044008801761e-06, "loss": 0.1435, "step": 4457 }, { "epoch": 1.38, "learning_rate": 8.219643928785758e-06, "loss": 0.1446, "step": 4458 }, { "epoch": 1.38, "learning_rate": 8.219243848769755e-06, "loss": 0.106, "step": 4459 }, { "epoch": 1.38, "learning_rate": 8.218843768753751e-06, "loss": 0.1724, "step": 4460 }, { "epoch": 1.38, "learning_rate": 8.218443688737748e-06, "loss": 0.1432, "step": 4461 }, { "epoch": 1.38, "learning_rate": 8.218043608721745e-06, "loss": 0.1175, "step": 4462 }, { "epoch": 1.39, "learning_rate": 8.217643528705741e-06, "loss": 0.2866, "step": 4463 }, { "epoch": 1.39, "learning_rate": 8.21724344868974e-06, "loss": 0.2003, "step": 4464 }, { "epoch": 1.39, "learning_rate": 8.216843368673736e-06, "loss": 0.1243, "step": 4465 }, { "epoch": 1.39, "learning_rate": 8.216443288657731e-06, "loss": 0.1429, "step": 4466 }, { "epoch": 1.39, "learning_rate": 8.216043208641728e-06, "loss": 0.1269, "step": 4467 }, { "epoch": 1.39, "learning_rate": 8.215643128625725e-06, "loss": 0.0778, "step": 4468 }, { "epoch": 1.39, "learning_rate": 8.215243048609723e-06, "loss": 0.1163, "step": 4469 }, { "epoch": 1.39, "learning_rate": 8.21484296859372e-06, "loss": 0.1602, "step": 4470 }, { "epoch": 1.39, "learning_rate": 8.214442888577717e-06, "loss": 0.1369, "step": 4471 }, { "epoch": 1.39, "learning_rate": 8.214042808561713e-06, "loss": 0.1981, "step": 4472 }, { "epoch": 1.39, "learning_rate": 8.21364272854571e-06, "loss": 0.1229, "step": 4473 }, { "epoch": 1.39, "learning_rate": 8.213242648529707e-06, "loss": 0.1023, "step": 4474 }, { "epoch": 1.39, "learning_rate": 8.212842568513703e-06, "loss": 0.123, "step": 4475 }, { "epoch": 1.39, "learning_rate": 8.2124424884977e-06, "loss": 0.3264, "step": 4476 }, { "epoch": 1.39, "learning_rate": 8.212042408481697e-06, "loss": 0.0947, "step": 4477 }, { "epoch": 1.39, "learning_rate": 8.211642328465695e-06, "loss": 0.1358, "step": 4478 }, { "epoch": 1.39, "learning_rate": 8.21124224844969e-06, "loss": 0.1492, "step": 4479 }, { "epoch": 1.39, "learning_rate": 8.210842168433687e-06, "loss": 0.1831, "step": 4480 }, { "epoch": 1.39, "learning_rate": 8.210442088417683e-06, "loss": 0.1685, "step": 4481 }, { "epoch": 1.39, "learning_rate": 8.21004200840168e-06, "loss": 0.2059, "step": 4482 }, { "epoch": 1.39, "learning_rate": 8.209641928385678e-06, "loss": 0.1617, "step": 4483 }, { "epoch": 1.39, "learning_rate": 8.209241848369675e-06, "loss": 0.125, "step": 4484 }, { "epoch": 1.39, "learning_rate": 8.208841768353672e-06, "loss": 0.1307, "step": 4485 }, { "epoch": 1.39, "learning_rate": 8.208441688337668e-06, "loss": 0.1161, "step": 4486 }, { "epoch": 1.39, "learning_rate": 8.208041608321665e-06, "loss": 0.0902, "step": 4487 }, { "epoch": 1.39, "learning_rate": 8.207641528305662e-06, "loss": 0.1442, "step": 4488 }, { "epoch": 1.39, "learning_rate": 8.207241448289659e-06, "loss": 0.1423, "step": 4489 }, { "epoch": 1.39, "learning_rate": 8.206841368273655e-06, "loss": 0.1771, "step": 4490 }, { "epoch": 1.39, "learning_rate": 8.206441288257652e-06, "loss": 0.2066, "step": 4491 }, { "epoch": 1.39, "learning_rate": 8.206041208241649e-06, "loss": 0.0795, "step": 4492 }, { "epoch": 1.39, "learning_rate": 8.205641128225645e-06, "loss": 0.1535, "step": 4493 }, { "epoch": 1.39, "learning_rate": 8.205241048209642e-06, "loss": 0.1294, "step": 4494 }, { "epoch": 1.4, "learning_rate": 8.204840968193639e-06, "loss": 0.231, "step": 4495 }, { "epoch": 1.4, "learning_rate": 8.204440888177635e-06, "loss": 0.1714, "step": 4496 }, { "epoch": 1.4, "learning_rate": 8.204040808161634e-06, "loss": 0.083, "step": 4497 }, { "epoch": 1.4, "learning_rate": 8.20364072814563e-06, "loss": 0.1317, "step": 4498 }, { "epoch": 1.4, "learning_rate": 8.203240648129627e-06, "loss": 0.0784, "step": 4499 }, { "epoch": 1.4, "learning_rate": 8.202840568113624e-06, "loss": 0.1293, "step": 4500 }, { "epoch": 1.4, "learning_rate": 8.20244048809762e-06, "loss": 0.1004, "step": 4501 }, { "epoch": 1.4, "learning_rate": 8.202040408081617e-06, "loss": 0.2426, "step": 4502 }, { "epoch": 1.4, "learning_rate": 8.201640328065614e-06, "loss": 0.1266, "step": 4503 }, { "epoch": 1.4, "learning_rate": 8.20124024804961e-06, "loss": 0.2517, "step": 4504 }, { "epoch": 1.4, "learning_rate": 8.200840168033607e-06, "loss": 0.2627, "step": 4505 }, { "epoch": 1.4, "learning_rate": 8.200440088017604e-06, "loss": 0.1971, "step": 4506 }, { "epoch": 1.4, "learning_rate": 8.2000400080016e-06, "loss": 0.1653, "step": 4507 }, { "epoch": 1.4, "learning_rate": 8.199639927985597e-06, "loss": 0.1925, "step": 4508 }, { "epoch": 1.4, "learning_rate": 8.199239847969594e-06, "loss": 0.2129, "step": 4509 }, { "epoch": 1.4, "learning_rate": 8.19883976795359e-06, "loss": 0.0711, "step": 4510 }, { "epoch": 1.4, "learning_rate": 8.198439687937589e-06, "loss": 0.2114, "step": 4511 }, { "epoch": 1.4, "learning_rate": 8.198039607921586e-06, "loss": 0.2209, "step": 4512 }, { "epoch": 1.4, "learning_rate": 8.197639527905582e-06, "loss": 0.1328, "step": 4513 }, { "epoch": 1.4, "learning_rate": 8.197239447889579e-06, "loss": 0.158, "step": 4514 }, { "epoch": 1.4, "learning_rate": 8.196839367873576e-06, "loss": 0.2418, "step": 4515 }, { "epoch": 1.4, "learning_rate": 8.196439287857572e-06, "loss": 0.1407, "step": 4516 }, { "epoch": 1.4, "learning_rate": 8.196039207841569e-06, "loss": 0.1424, "step": 4517 }, { "epoch": 1.4, "learning_rate": 8.195639127825566e-06, "loss": 0.1946, "step": 4518 }, { "epoch": 1.4, "learning_rate": 8.195239047809563e-06, "loss": 0.2679, "step": 4519 }, { "epoch": 1.4, "learning_rate": 8.19483896779356e-06, "loss": 0.155, "step": 4520 }, { "epoch": 1.4, "learning_rate": 8.194438887777556e-06, "loss": 0.1749, "step": 4521 }, { "epoch": 1.4, "learning_rate": 8.194038807761553e-06, "loss": 0.0773, "step": 4522 }, { "epoch": 1.4, "learning_rate": 8.19363872774555e-06, "loss": 0.2923, "step": 4523 }, { "epoch": 1.4, "learning_rate": 8.193238647729546e-06, "loss": 0.2051, "step": 4524 }, { "epoch": 1.4, "learning_rate": 8.192838567713544e-06, "loss": 0.0898, "step": 4525 }, { "epoch": 1.4, "learning_rate": 8.192438487697541e-06, "loss": 0.1439, "step": 4526 }, { "epoch": 1.41, "learning_rate": 8.192038407681538e-06, "loss": 0.1548, "step": 4527 }, { "epoch": 1.41, "learning_rate": 8.191638327665534e-06, "loss": 0.1057, "step": 4528 }, { "epoch": 1.41, "learning_rate": 8.191238247649531e-06, "loss": 0.1538, "step": 4529 }, { "epoch": 1.41, "learning_rate": 8.190838167633526e-06, "loss": 0.1313, "step": 4530 }, { "epoch": 1.41, "learning_rate": 8.190438087617524e-06, "loss": 0.2127, "step": 4531 }, { "epoch": 1.41, "learning_rate": 8.190038007601521e-06, "loss": 0.0814, "step": 4532 }, { "epoch": 1.41, "learning_rate": 8.189637927585518e-06, "loss": 0.1014, "step": 4533 }, { "epoch": 1.41, "learning_rate": 8.189237847569515e-06, "loss": 0.1521, "step": 4534 }, { "epoch": 1.41, "learning_rate": 8.188837767553511e-06, "loss": 0.1627, "step": 4535 }, { "epoch": 1.41, "learning_rate": 8.188437687537508e-06, "loss": 0.1653, "step": 4536 }, { "epoch": 1.41, "learning_rate": 8.188037607521505e-06, "loss": 0.086, "step": 4537 }, { "epoch": 1.41, "learning_rate": 8.187637527505501e-06, "loss": 0.1346, "step": 4538 }, { "epoch": 1.41, "learning_rate": 8.1872374474895e-06, "loss": 0.1249, "step": 4539 }, { "epoch": 1.41, "learning_rate": 8.186837367473496e-06, "loss": 0.152, "step": 4540 }, { "epoch": 1.41, "learning_rate": 8.186437287457493e-06, "loss": 0.1456, "step": 4541 }, { "epoch": 1.41, "learning_rate": 8.186037207441488e-06, "loss": 0.1265, "step": 4542 }, { "epoch": 1.41, "learning_rate": 8.185637127425485e-06, "loss": 0.1023, "step": 4543 }, { "epoch": 1.41, "learning_rate": 8.185237047409481e-06, "loss": 0.1219, "step": 4544 }, { "epoch": 1.41, "learning_rate": 8.18483696739348e-06, "loss": 0.1233, "step": 4545 }, { "epoch": 1.41, "learning_rate": 8.184436887377476e-06, "loss": 0.2057, "step": 4546 }, { "epoch": 1.41, "learning_rate": 8.184036807361473e-06, "loss": 0.1559, "step": 4547 }, { "epoch": 1.41, "learning_rate": 8.18363672734547e-06, "loss": 0.2133, "step": 4548 }, { "epoch": 1.41, "learning_rate": 8.183236647329467e-06, "loss": 0.1323, "step": 4549 }, { "epoch": 1.41, "learning_rate": 8.182836567313463e-06, "loss": 0.1523, "step": 4550 }, { "epoch": 1.41, "learning_rate": 8.18243648729746e-06, "loss": 0.153, "step": 4551 }, { "epoch": 1.41, "learning_rate": 8.182036407281457e-06, "loss": 0.1687, "step": 4552 }, { "epoch": 1.41, "learning_rate": 8.181636327265455e-06, "loss": 0.137, "step": 4553 }, { "epoch": 1.41, "learning_rate": 8.181236247249452e-06, "loss": 0.1673, "step": 4554 }, { "epoch": 1.41, "learning_rate": 8.180836167233447e-06, "loss": 0.1329, "step": 4555 }, { "epoch": 1.41, "learning_rate": 8.180436087217443e-06, "loss": 0.1373, "step": 4556 }, { "epoch": 1.41, "learning_rate": 8.18003600720144e-06, "loss": 0.2124, "step": 4557 }, { "epoch": 1.41, "learning_rate": 8.179635927185438e-06, "loss": 0.1809, "step": 4558 }, { "epoch": 1.42, "learning_rate": 8.179235847169435e-06, "loss": 0.1475, "step": 4559 }, { "epoch": 1.42, "learning_rate": 8.178835767153432e-06, "loss": 0.2811, "step": 4560 }, { "epoch": 1.42, "learning_rate": 8.178435687137428e-06, "loss": 0.0783, "step": 4561 }, { "epoch": 1.42, "learning_rate": 8.178035607121425e-06, "loss": 0.1139, "step": 4562 }, { "epoch": 1.42, "learning_rate": 8.177635527105422e-06, "loss": 0.1249, "step": 4563 }, { "epoch": 1.42, "learning_rate": 8.177235447089418e-06, "loss": 0.1646, "step": 4564 }, { "epoch": 1.42, "learning_rate": 8.176835367073415e-06, "loss": 0.1596, "step": 4565 }, { "epoch": 1.42, "learning_rate": 8.176435287057412e-06, "loss": 0.1146, "step": 4566 }, { "epoch": 1.42, "learning_rate": 8.176035207041409e-06, "loss": 0.0808, "step": 4567 }, { "epoch": 1.42, "learning_rate": 8.175635127025405e-06, "loss": 0.1597, "step": 4568 }, { "epoch": 1.42, "learning_rate": 8.175235047009402e-06, "loss": 0.2077, "step": 4569 }, { "epoch": 1.42, "learning_rate": 8.174834966993399e-06, "loss": 0.1266, "step": 4570 }, { "epoch": 1.42, "learning_rate": 8.174434886977395e-06, "loss": 0.3483, "step": 4571 }, { "epoch": 1.42, "learning_rate": 8.174034806961394e-06, "loss": 0.174, "step": 4572 }, { "epoch": 1.42, "learning_rate": 8.17363472694539e-06, "loss": 0.1156, "step": 4573 }, { "epoch": 1.42, "learning_rate": 8.173234646929387e-06, "loss": 0.1486, "step": 4574 }, { "epoch": 1.42, "learning_rate": 8.172834566913384e-06, "loss": 0.1306, "step": 4575 }, { "epoch": 1.42, "learning_rate": 8.17243448689738e-06, "loss": 0.0585, "step": 4576 }, { "epoch": 1.42, "learning_rate": 8.172034406881377e-06, "loss": 0.1029, "step": 4577 }, { "epoch": 1.42, "learning_rate": 8.171634326865374e-06, "loss": 0.1693, "step": 4578 }, { "epoch": 1.42, "learning_rate": 8.17123424684937e-06, "loss": 0.1767, "step": 4579 }, { "epoch": 1.42, "learning_rate": 8.170834166833367e-06, "loss": 0.2166, "step": 4580 }, { "epoch": 1.42, "learning_rate": 8.170434086817364e-06, "loss": 0.172, "step": 4581 }, { "epoch": 1.42, "learning_rate": 8.17003400680136e-06, "loss": 0.1748, "step": 4582 }, { "epoch": 1.42, "learning_rate": 8.169633926785357e-06, "loss": 0.0745, "step": 4583 }, { "epoch": 1.42, "learning_rate": 8.169233846769354e-06, "loss": 0.1394, "step": 4584 }, { "epoch": 1.42, "learning_rate": 8.16883376675335e-06, "loss": 0.1354, "step": 4585 }, { "epoch": 1.42, "learning_rate": 8.168433686737349e-06, "loss": 0.1201, "step": 4586 }, { "epoch": 1.42, "learning_rate": 8.168033606721346e-06, "loss": 0.1243, "step": 4587 }, { "epoch": 1.42, "learning_rate": 8.167633526705342e-06, "loss": 0.1836, "step": 4588 }, { "epoch": 1.42, "learning_rate": 8.167233446689339e-06, "loss": 0.1723, "step": 4589 }, { "epoch": 1.42, "learning_rate": 8.166833366673336e-06, "loss": 0.1359, "step": 4590 }, { "epoch": 1.42, "learning_rate": 8.166433286657332e-06, "loss": 0.1005, "step": 4591 }, { "epoch": 1.43, "learning_rate": 8.166033206641329e-06, "loss": 0.0993, "step": 4592 }, { "epoch": 1.43, "learning_rate": 8.165633126625326e-06, "loss": 0.0951, "step": 4593 }, { "epoch": 1.43, "learning_rate": 8.165233046609322e-06, "loss": 0.1141, "step": 4594 }, { "epoch": 1.43, "learning_rate": 8.164832966593319e-06, "loss": 0.1893, "step": 4595 }, { "epoch": 1.43, "learning_rate": 8.164432886577316e-06, "loss": 0.108, "step": 4596 }, { "epoch": 1.43, "learning_rate": 8.164032806561313e-06, "loss": 0.1884, "step": 4597 }, { "epoch": 1.43, "learning_rate": 8.16363272654531e-06, "loss": 0.1586, "step": 4598 }, { "epoch": 1.43, "learning_rate": 8.163232646529306e-06, "loss": 0.1883, "step": 4599 }, { "epoch": 1.43, "learning_rate": 8.162832566513304e-06, "loss": 0.0928, "step": 4600 }, { "epoch": 1.43, "learning_rate": 8.162432486497301e-06, "loss": 0.0755, "step": 4601 }, { "epoch": 1.43, "learning_rate": 8.162032406481298e-06, "loss": 0.1536, "step": 4602 }, { "epoch": 1.43, "learning_rate": 8.161632326465294e-06, "loss": 0.1374, "step": 4603 }, { "epoch": 1.43, "learning_rate": 8.161232246449291e-06, "loss": 0.1805, "step": 4604 }, { "epoch": 1.43, "learning_rate": 8.160832166433286e-06, "loss": 0.1351, "step": 4605 }, { "epoch": 1.43, "learning_rate": 8.160432086417284e-06, "loss": 0.1373, "step": 4606 }, { "epoch": 1.43, "learning_rate": 8.160032006401281e-06, "loss": 0.1191, "step": 4607 }, { "epoch": 1.43, "learning_rate": 8.159631926385278e-06, "loss": 0.2345, "step": 4608 }, { "epoch": 1.43, "learning_rate": 8.159231846369274e-06, "loss": 0.2103, "step": 4609 }, { "epoch": 1.43, "learning_rate": 8.158831766353271e-06, "loss": 0.097, "step": 4610 }, { "epoch": 1.43, "learning_rate": 8.158431686337268e-06, "loss": 0.07, "step": 4611 }, { "epoch": 1.43, "learning_rate": 8.158031606321265e-06, "loss": 0.2133, "step": 4612 }, { "epoch": 1.43, "learning_rate": 8.157631526305261e-06, "loss": 0.2331, "step": 4613 }, { "epoch": 1.43, "learning_rate": 8.15723144628926e-06, "loss": 0.11, "step": 4614 }, { "epoch": 1.43, "learning_rate": 8.156831366273256e-06, "loss": 0.0998, "step": 4615 }, { "epoch": 1.43, "learning_rate": 8.156431286257253e-06, "loss": 0.1774, "step": 4616 }, { "epoch": 1.43, "learning_rate": 8.15603120624125e-06, "loss": 0.154, "step": 4617 }, { "epoch": 1.43, "learning_rate": 8.155631126225245e-06, "loss": 0.1686, "step": 4618 }, { "epoch": 1.43, "learning_rate": 8.155231046209241e-06, "loss": 0.1912, "step": 4619 }, { "epoch": 1.43, "learning_rate": 8.15483096619324e-06, "loss": 0.1164, "step": 4620 }, { "epoch": 1.43, "learning_rate": 8.154430886177236e-06, "loss": 0.2597, "step": 4621 }, { "epoch": 1.43, "learning_rate": 8.154030806161233e-06, "loss": 0.1506, "step": 4622 }, { "epoch": 1.43, "learning_rate": 8.15363072614523e-06, "loss": 0.1798, "step": 4623 }, { "epoch": 1.44, "learning_rate": 8.153230646129226e-06, "loss": 0.1646, "step": 4624 }, { "epoch": 1.44, "learning_rate": 8.152830566113223e-06, "loss": 0.1892, "step": 4625 }, { "epoch": 1.44, "learning_rate": 8.15243048609722e-06, "loss": 0.1989, "step": 4626 }, { "epoch": 1.44, "learning_rate": 8.152030406081217e-06, "loss": 0.198, "step": 4627 }, { "epoch": 1.44, "learning_rate": 8.151630326065215e-06, "loss": 0.2152, "step": 4628 }, { "epoch": 1.44, "learning_rate": 8.151230246049212e-06, "loss": 0.1468, "step": 4629 }, { "epoch": 1.44, "learning_rate": 8.150830166033207e-06, "loss": 0.141, "step": 4630 }, { "epoch": 1.44, "learning_rate": 8.150430086017203e-06, "loss": 0.0467, "step": 4631 }, { "epoch": 1.44, "learning_rate": 8.1500300060012e-06, "loss": 0.2425, "step": 4632 }, { "epoch": 1.44, "learning_rate": 8.149629925985197e-06, "loss": 0.1232, "step": 4633 }, { "epoch": 1.44, "learning_rate": 8.149229845969195e-06, "loss": 0.2223, "step": 4634 }, { "epoch": 1.44, "learning_rate": 8.148829765953192e-06, "loss": 0.1263, "step": 4635 }, { "epoch": 1.44, "learning_rate": 8.148429685937188e-06, "loss": 0.1739, "step": 4636 }, { "epoch": 1.44, "learning_rate": 8.148029605921185e-06, "loss": 0.1372, "step": 4637 }, { "epoch": 1.44, "learning_rate": 8.147629525905182e-06, "loss": 0.1628, "step": 4638 }, { "epoch": 1.44, "learning_rate": 8.147229445889178e-06, "loss": 0.1281, "step": 4639 }, { "epoch": 1.44, "learning_rate": 8.146829365873175e-06, "loss": 0.1622, "step": 4640 }, { "epoch": 1.44, "learning_rate": 8.146429285857172e-06, "loss": 0.173, "step": 4641 }, { "epoch": 1.44, "learning_rate": 8.14602920584117e-06, "loss": 0.2092, "step": 4642 }, { "epoch": 1.44, "learning_rate": 8.145629125825165e-06, "loss": 0.1957, "step": 4643 }, { "epoch": 1.44, "learning_rate": 8.145229045809162e-06, "loss": 0.1967, "step": 4644 }, { "epoch": 1.44, "learning_rate": 8.144828965793159e-06, "loss": 0.3302, "step": 4645 }, { "epoch": 1.44, "learning_rate": 8.144428885777155e-06, "loss": 0.1922, "step": 4646 }, { "epoch": 1.44, "learning_rate": 8.144028805761152e-06, "loss": 0.0992, "step": 4647 }, { "epoch": 1.44, "learning_rate": 8.14362872574515e-06, "loss": 0.0555, "step": 4648 }, { "epoch": 1.44, "learning_rate": 8.143228645729147e-06, "loss": 0.1415, "step": 4649 }, { "epoch": 1.44, "learning_rate": 8.142828565713144e-06, "loss": 0.2052, "step": 4650 }, { "epoch": 1.44, "learning_rate": 8.14242848569714e-06, "loss": 0.1928, "step": 4651 }, { "epoch": 1.44, "learning_rate": 8.142028405681137e-06, "loss": 0.2013, "step": 4652 }, { "epoch": 1.44, "learning_rate": 8.141628325665134e-06, "loss": 0.2011, "step": 4653 }, { "epoch": 1.44, "learning_rate": 8.14122824564913e-06, "loss": 0.1255, "step": 4654 }, { "epoch": 1.44, "learning_rate": 8.140828165633127e-06, "loss": 0.163, "step": 4655 }, { "epoch": 1.45, "learning_rate": 8.140428085617124e-06, "loss": 0.1613, "step": 4656 }, { "epoch": 1.45, "learning_rate": 8.14002800560112e-06, "loss": 0.2159, "step": 4657 }, { "epoch": 1.45, "learning_rate": 8.139627925585117e-06, "loss": 0.157, "step": 4658 }, { "epoch": 1.45, "learning_rate": 8.139227845569114e-06, "loss": 0.1519, "step": 4659 }, { "epoch": 1.45, "learning_rate": 8.13882776555311e-06, "loss": 0.2448, "step": 4660 }, { "epoch": 1.45, "learning_rate": 8.138427685537107e-06, "loss": 0.2106, "step": 4661 }, { "epoch": 1.45, "learning_rate": 8.138027605521106e-06, "loss": 0.1356, "step": 4662 }, { "epoch": 1.45, "learning_rate": 8.137627525505102e-06, "loss": 0.0802, "step": 4663 }, { "epoch": 1.45, "learning_rate": 8.137227445489099e-06, "loss": 0.1593, "step": 4664 }, { "epoch": 1.45, "learning_rate": 8.136827365473096e-06, "loss": 0.2686, "step": 4665 }, { "epoch": 1.45, "learning_rate": 8.136427285457092e-06, "loss": 0.1511, "step": 4666 }, { "epoch": 1.45, "learning_rate": 8.136027205441089e-06, "loss": 0.1279, "step": 4667 }, { "epoch": 1.45, "learning_rate": 8.135627125425086e-06, "loss": 0.151, "step": 4668 }, { "epoch": 1.45, "learning_rate": 8.135227045409082e-06, "loss": 0.1302, "step": 4669 }, { "epoch": 1.45, "learning_rate": 8.134826965393079e-06, "loss": 0.1384, "step": 4670 }, { "epoch": 1.45, "learning_rate": 8.134426885377076e-06, "loss": 0.1052, "step": 4671 }, { "epoch": 1.45, "learning_rate": 8.134026805361072e-06, "loss": 0.1749, "step": 4672 }, { "epoch": 1.45, "learning_rate": 8.133626725345069e-06, "loss": 0.1108, "step": 4673 }, { "epoch": 1.45, "learning_rate": 8.133226645329066e-06, "loss": 0.2307, "step": 4674 }, { "epoch": 1.45, "learning_rate": 8.132826565313063e-06, "loss": 0.2312, "step": 4675 }, { "epoch": 1.45, "learning_rate": 8.132426485297061e-06, "loss": 0.1592, "step": 4676 }, { "epoch": 1.45, "learning_rate": 8.132026405281058e-06, "loss": 0.0887, "step": 4677 }, { "epoch": 1.45, "learning_rate": 8.131626325265054e-06, "loss": 0.1258, "step": 4678 }, { "epoch": 1.45, "learning_rate": 8.131226245249051e-06, "loss": 0.2382, "step": 4679 }, { "epoch": 1.45, "learning_rate": 8.130826165233048e-06, "loss": 0.1376, "step": 4680 }, { "epoch": 1.45, "learning_rate": 8.130426085217044e-06, "loss": 0.2209, "step": 4681 }, { "epoch": 1.45, "learning_rate": 8.130026005201041e-06, "loss": 0.2868, "step": 4682 }, { "epoch": 1.45, "learning_rate": 8.129625925185038e-06, "loss": 0.1109, "step": 4683 }, { "epoch": 1.45, "learning_rate": 8.129225845169034e-06, "loss": 0.148, "step": 4684 }, { "epoch": 1.45, "learning_rate": 8.128825765153031e-06, "loss": 0.1863, "step": 4685 }, { "epoch": 1.45, "learning_rate": 8.128425685137028e-06, "loss": 0.1672, "step": 4686 }, { "epoch": 1.45, "learning_rate": 8.128025605121024e-06, "loss": 0.1626, "step": 4687 }, { "epoch": 1.46, "learning_rate": 8.127625525105021e-06, "loss": 0.2853, "step": 4688 }, { "epoch": 1.46, "learning_rate": 8.127225445089018e-06, "loss": 0.1618, "step": 4689 }, { "epoch": 1.46, "learning_rate": 8.126825365073016e-06, "loss": 0.1417, "step": 4690 }, { "epoch": 1.46, "learning_rate": 8.126425285057013e-06, "loss": 0.1728, "step": 4691 }, { "epoch": 1.46, "learning_rate": 8.12602520504101e-06, "loss": 0.109, "step": 4692 }, { "epoch": 1.46, "learning_rate": 8.125625125025005e-06, "loss": 0.1313, "step": 4693 }, { "epoch": 1.46, "learning_rate": 8.125225045009001e-06, "loss": 0.1063, "step": 4694 }, { "epoch": 1.46, "learning_rate": 8.124824964993e-06, "loss": 0.1635, "step": 4695 }, { "epoch": 1.46, "learning_rate": 8.124424884976996e-06, "loss": 0.1212, "step": 4696 }, { "epoch": 1.46, "learning_rate": 8.124024804960993e-06, "loss": 0.2025, "step": 4697 }, { "epoch": 1.46, "learning_rate": 8.12362472494499e-06, "loss": 0.0901, "step": 4698 }, { "epoch": 1.46, "learning_rate": 8.123224644928986e-06, "loss": 0.1241, "step": 4699 }, { "epoch": 1.46, "learning_rate": 8.122824564912983e-06, "loss": 0.1264, "step": 4700 }, { "epoch": 1.46, "learning_rate": 8.12242448489698e-06, "loss": 0.1746, "step": 4701 }, { "epoch": 1.46, "learning_rate": 8.122024404880976e-06, "loss": 0.1702, "step": 4702 }, { "epoch": 1.46, "learning_rate": 8.121624324864975e-06, "loss": 0.1702, "step": 4703 }, { "epoch": 1.46, "learning_rate": 8.121224244848972e-06, "loss": 0.1867, "step": 4704 }, { "epoch": 1.46, "learning_rate": 8.120824164832968e-06, "loss": 0.2628, "step": 4705 }, { "epoch": 1.46, "learning_rate": 8.120424084816963e-06, "loss": 0.1187, "step": 4706 }, { "epoch": 1.46, "learning_rate": 8.12002400480096e-06, "loss": 0.1001, "step": 4707 }, { "epoch": 1.46, "learning_rate": 8.119623924784957e-06, "loss": 0.2793, "step": 4708 }, { "epoch": 1.46, "learning_rate": 8.119223844768955e-06, "loss": 0.1873, "step": 4709 }, { "epoch": 1.46, "learning_rate": 8.118823764752952e-06, "loss": 0.1211, "step": 4710 }, { "epoch": 1.46, "learning_rate": 8.118423684736948e-06, "loss": 0.1345, "step": 4711 }, { "epoch": 1.46, "learning_rate": 8.118023604720945e-06, "loss": 0.35, "step": 4712 }, { "epoch": 1.46, "learning_rate": 8.117623524704942e-06, "loss": 0.1801, "step": 4713 }, { "epoch": 1.46, "learning_rate": 8.117223444688938e-06, "loss": 0.1791, "step": 4714 }, { "epoch": 1.46, "learning_rate": 8.116823364672935e-06, "loss": 0.1628, "step": 4715 }, { "epoch": 1.46, "learning_rate": 8.116423284656932e-06, "loss": 0.1454, "step": 4716 }, { "epoch": 1.46, "learning_rate": 8.11602320464093e-06, "loss": 0.2647, "step": 4717 }, { "epoch": 1.46, "learning_rate": 8.115623124624927e-06, "loss": 0.1205, "step": 4718 }, { "epoch": 1.46, "learning_rate": 8.115223044608922e-06, "loss": 0.2151, "step": 4719 }, { "epoch": 1.46, "learning_rate": 8.114822964592918e-06, "loss": 0.1279, "step": 4720 }, { "epoch": 1.47, "learning_rate": 8.114422884576915e-06, "loss": 0.2055, "step": 4721 }, { "epoch": 1.47, "learning_rate": 8.114022804560912e-06, "loss": 0.2497, "step": 4722 }, { "epoch": 1.47, "learning_rate": 8.11362272454491e-06, "loss": 0.1398, "step": 4723 }, { "epoch": 1.47, "learning_rate": 8.113222644528907e-06, "loss": 0.169, "step": 4724 }, { "epoch": 1.47, "learning_rate": 8.112822564512904e-06, "loss": 0.1251, "step": 4725 }, { "epoch": 1.47, "learning_rate": 8.1124224844969e-06, "loss": 0.2503, "step": 4726 }, { "epoch": 1.47, "learning_rate": 8.112022404480897e-06, "loss": 0.1295, "step": 4727 }, { "epoch": 1.47, "learning_rate": 8.111622324464894e-06, "loss": 0.0918, "step": 4728 }, { "epoch": 1.47, "learning_rate": 8.11122224444889e-06, "loss": 0.1056, "step": 4729 }, { "epoch": 1.47, "learning_rate": 8.110822164432887e-06, "loss": 0.0964, "step": 4730 }, { "epoch": 1.47, "learning_rate": 8.110422084416884e-06, "loss": 0.1851, "step": 4731 }, { "epoch": 1.47, "learning_rate": 8.11002200440088e-06, "loss": 0.1046, "step": 4732 }, { "epoch": 1.47, "learning_rate": 8.109621924384877e-06, "loss": 0.1956, "step": 4733 }, { "epoch": 1.47, "learning_rate": 8.109221844368874e-06, "loss": 0.1499, "step": 4734 }, { "epoch": 1.47, "learning_rate": 8.10882176435287e-06, "loss": 0.1024, "step": 4735 }, { "epoch": 1.47, "learning_rate": 8.108421684336867e-06, "loss": 0.1295, "step": 4736 }, { "epoch": 1.47, "learning_rate": 8.108021604320866e-06, "loss": 0.1629, "step": 4737 }, { "epoch": 1.47, "learning_rate": 8.107621524304862e-06, "loss": 0.2234, "step": 4738 }, { "epoch": 1.47, "learning_rate": 8.107221444288859e-06, "loss": 0.1414, "step": 4739 }, { "epoch": 1.47, "learning_rate": 8.106821364272856e-06, "loss": 0.1971, "step": 4740 }, { "epoch": 1.47, "learning_rate": 8.106421284256852e-06, "loss": 0.3993, "step": 4741 }, { "epoch": 1.47, "learning_rate": 8.106021204240849e-06, "loss": 0.1628, "step": 4742 }, { "epoch": 1.47, "learning_rate": 8.105621124224846e-06, "loss": 0.2717, "step": 4743 }, { "epoch": 1.47, "learning_rate": 8.105221044208842e-06, "loss": 0.1186, "step": 4744 }, { "epoch": 1.47, "learning_rate": 8.104820964192839e-06, "loss": 0.2987, "step": 4745 }, { "epoch": 1.47, "learning_rate": 8.104420884176836e-06, "loss": 0.2337, "step": 4746 }, { "epoch": 1.47, "learning_rate": 8.104020804160832e-06, "loss": 0.2126, "step": 4747 }, { "epoch": 1.47, "learning_rate": 8.103620724144829e-06, "loss": 0.1079, "step": 4748 }, { "epoch": 1.47, "learning_rate": 8.103220644128826e-06, "loss": 0.2894, "step": 4749 }, { "epoch": 1.47, "learning_rate": 8.102820564112822e-06, "loss": 0.1299, "step": 4750 }, { "epoch": 1.47, "learning_rate": 8.10242048409682e-06, "loss": 0.1521, "step": 4751 }, { "epoch": 1.47, "learning_rate": 8.102020404080818e-06, "loss": 0.1298, "step": 4752 }, { "epoch": 1.48, "learning_rate": 8.101620324064814e-06, "loss": 0.1001, "step": 4753 }, { "epoch": 1.48, "learning_rate": 8.101220244048811e-06, "loss": 0.2743, "step": 4754 }, { "epoch": 1.48, "learning_rate": 8.100820164032808e-06, "loss": 0.1113, "step": 4755 }, { "epoch": 1.48, "learning_rate": 8.100420084016804e-06, "loss": 0.1332, "step": 4756 }, { "epoch": 1.48, "learning_rate": 8.100020004000801e-06, "loss": 0.1667, "step": 4757 }, { "epoch": 1.48, "learning_rate": 8.099619923984798e-06, "loss": 0.1021, "step": 4758 }, { "epoch": 1.48, "learning_rate": 8.099219843968794e-06, "loss": 0.1123, "step": 4759 }, { "epoch": 1.48, "learning_rate": 8.098819763952791e-06, "loss": 0.1012, "step": 4760 }, { "epoch": 1.48, "learning_rate": 8.098419683936788e-06, "loss": 0.126, "step": 4761 }, { "epoch": 1.48, "learning_rate": 8.098019603920784e-06, "loss": 0.134, "step": 4762 }, { "epoch": 1.48, "learning_rate": 8.097619523904781e-06, "loss": 0.1352, "step": 4763 }, { "epoch": 1.48, "learning_rate": 8.097219443888778e-06, "loss": 0.1038, "step": 4764 }, { "epoch": 1.48, "learning_rate": 8.096819363872776e-06, "loss": 0.0887, "step": 4765 }, { "epoch": 1.48, "learning_rate": 8.096419283856773e-06, "loss": 0.3448, "step": 4766 }, { "epoch": 1.48, "learning_rate": 8.09601920384077e-06, "loss": 0.1421, "step": 4767 }, { "epoch": 1.48, "learning_rate": 8.095619123824766e-06, "loss": 0.2128, "step": 4768 }, { "epoch": 1.48, "learning_rate": 8.095219043808761e-06, "loss": 0.0807, "step": 4769 }, { "epoch": 1.48, "learning_rate": 8.09481896379276e-06, "loss": 0.2219, "step": 4770 }, { "epoch": 1.48, "learning_rate": 8.094418883776756e-06, "loss": 0.1192, "step": 4771 }, { "epoch": 1.48, "learning_rate": 8.094018803760753e-06, "loss": 0.2273, "step": 4772 }, { "epoch": 1.48, "learning_rate": 8.09361872374475e-06, "loss": 0.1209, "step": 4773 }, { "epoch": 1.48, "learning_rate": 8.093218643728746e-06, "loss": 0.1744, "step": 4774 }, { "epoch": 1.48, "learning_rate": 8.092818563712743e-06, "loss": 0.1474, "step": 4775 }, { "epoch": 1.48, "learning_rate": 8.09241848369674e-06, "loss": 0.1176, "step": 4776 }, { "epoch": 1.48, "learning_rate": 8.092018403680736e-06, "loss": 0.0883, "step": 4777 }, { "epoch": 1.48, "learning_rate": 8.091618323664733e-06, "loss": 0.1365, "step": 4778 }, { "epoch": 1.48, "learning_rate": 8.091218243648731e-06, "loss": 0.1272, "step": 4779 }, { "epoch": 1.48, "learning_rate": 8.090818163632728e-06, "loss": 0.17, "step": 4780 }, { "epoch": 1.48, "learning_rate": 8.090418083616725e-06, "loss": 0.2131, "step": 4781 }, { "epoch": 1.48, "learning_rate": 8.09001800360072e-06, "loss": 0.1664, "step": 4782 }, { "epoch": 1.48, "learning_rate": 8.089617923584717e-06, "loss": 0.239, "step": 4783 }, { "epoch": 1.48, "learning_rate": 8.089217843568715e-06, "loss": 0.0993, "step": 4784 }, { "epoch": 1.49, "learning_rate": 8.088817763552712e-06, "loss": 0.1114, "step": 4785 }, { "epoch": 1.49, "learning_rate": 8.088417683536708e-06, "loss": 0.1955, "step": 4786 }, { "epoch": 1.49, "learning_rate": 8.088017603520705e-06, "loss": 0.1242, "step": 4787 }, { "epoch": 1.49, "learning_rate": 8.087617523504702e-06, "loss": 0.1091, "step": 4788 }, { "epoch": 1.49, "learning_rate": 8.087217443488698e-06, "loss": 0.1322, "step": 4789 }, { "epoch": 1.49, "learning_rate": 8.086817363472695e-06, "loss": 0.0797, "step": 4790 }, { "epoch": 1.49, "learning_rate": 8.086417283456692e-06, "loss": 0.199, "step": 4791 }, { "epoch": 1.49, "learning_rate": 8.086017203440688e-06, "loss": 0.1243, "step": 4792 }, { "epoch": 1.49, "learning_rate": 8.085617123424687e-06, "loss": 0.1459, "step": 4793 }, { "epoch": 1.49, "learning_rate": 8.085217043408682e-06, "loss": 0.075, "step": 4794 }, { "epoch": 1.49, "learning_rate": 8.084816963392678e-06, "loss": 0.0813, "step": 4795 }, { "epoch": 1.49, "learning_rate": 8.084416883376675e-06, "loss": 0.2814, "step": 4796 }, { "epoch": 1.49, "learning_rate": 8.084016803360672e-06, "loss": 0.3422, "step": 4797 }, { "epoch": 1.49, "learning_rate": 8.08361672334467e-06, "loss": 0.2146, "step": 4798 }, { "epoch": 1.49, "learning_rate": 8.083216643328667e-06, "loss": 0.1197, "step": 4799 }, { "epoch": 1.49, "learning_rate": 8.082816563312664e-06, "loss": 0.285, "step": 4800 }, { "epoch": 1.49, "learning_rate": 8.08241648329666e-06, "loss": 0.1441, "step": 4801 }, { "epoch": 1.49, "learning_rate": 8.082016403280657e-06, "loss": 0.2464, "step": 4802 }, { "epoch": 1.49, "learning_rate": 8.081616323264654e-06, "loss": 0.2344, "step": 4803 }, { "epoch": 1.49, "learning_rate": 8.08121624324865e-06, "loss": 0.1044, "step": 4804 }, { "epoch": 1.49, "learning_rate": 8.080816163232647e-06, "loss": 0.1395, "step": 4805 }, { "epoch": 1.49, "learning_rate": 8.080416083216644e-06, "loss": 0.1914, "step": 4806 }, { "epoch": 1.49, "learning_rate": 8.08001600320064e-06, "loss": 0.0959, "step": 4807 }, { "epoch": 1.49, "learning_rate": 8.079615923184637e-06, "loss": 0.1272, "step": 4808 }, { "epoch": 1.49, "learning_rate": 8.079215843168634e-06, "loss": 0.189, "step": 4809 }, { "epoch": 1.49, "learning_rate": 8.07881576315263e-06, "loss": 0.0969, "step": 4810 }, { "epoch": 1.49, "learning_rate": 8.078415683136627e-06, "loss": 0.098, "step": 4811 }, { "epoch": 1.49, "learning_rate": 8.078015603120625e-06, "loss": 0.215, "step": 4812 }, { "epoch": 1.49, "learning_rate": 8.077615523104622e-06, "loss": 0.3561, "step": 4813 }, { "epoch": 1.49, "learning_rate": 8.077215443088619e-06, "loss": 0.166, "step": 4814 }, { "epoch": 1.49, "learning_rate": 8.076815363072616e-06, "loss": 0.1231, "step": 4815 }, { "epoch": 1.49, "learning_rate": 8.076415283056612e-06, "loss": 0.1599, "step": 4816 }, { "epoch": 1.5, "learning_rate": 8.076015203040609e-06, "loss": 0.0863, "step": 4817 }, { "epoch": 1.5, "learning_rate": 8.075615123024606e-06, "loss": 0.1942, "step": 4818 }, { "epoch": 1.5, "learning_rate": 8.075215043008602e-06, "loss": 0.1897, "step": 4819 }, { "epoch": 1.5, "learning_rate": 8.074814962992599e-06, "loss": 0.1216, "step": 4820 }, { "epoch": 1.5, "learning_rate": 8.074414882976596e-06, "loss": 0.2006, "step": 4821 }, { "epoch": 1.5, "learning_rate": 8.074014802960592e-06, "loss": 0.2185, "step": 4822 }, { "epoch": 1.5, "learning_rate": 8.073614722944589e-06, "loss": 0.1508, "step": 4823 }, { "epoch": 1.5, "learning_rate": 8.073214642928586e-06, "loss": 0.1655, "step": 4824 }, { "epoch": 1.5, "learning_rate": 8.072814562912582e-06, "loss": 0.1041, "step": 4825 }, { "epoch": 1.5, "learning_rate": 8.07241448289658e-06, "loss": 0.2319, "step": 4826 }, { "epoch": 1.5, "learning_rate": 8.072014402880577e-06, "loss": 0.1205, "step": 4827 }, { "epoch": 1.5, "learning_rate": 8.071614322864574e-06, "loss": 0.1054, "step": 4828 }, { "epoch": 1.5, "learning_rate": 8.07121424284857e-06, "loss": 0.234, "step": 4829 }, { "epoch": 1.5, "learning_rate": 8.070814162832568e-06, "loss": 0.1417, "step": 4830 }, { "epoch": 1.5, "learning_rate": 8.070414082816564e-06, "loss": 0.1581, "step": 4831 }, { "epoch": 1.5, "learning_rate": 8.070014002800561e-06, "loss": 0.072, "step": 4832 }, { "epoch": 1.5, "learning_rate": 8.069613922784558e-06, "loss": 0.1232, "step": 4833 }, { "epoch": 1.5, "learning_rate": 8.069213842768554e-06, "loss": 0.1261, "step": 4834 }, { "epoch": 1.5, "learning_rate": 8.068813762752551e-06, "loss": 0.176, "step": 4835 }, { "epoch": 1.5, "learning_rate": 8.068413682736548e-06, "loss": 0.1131, "step": 4836 }, { "epoch": 1.5, "learning_rate": 8.068013602720544e-06, "loss": 0.216, "step": 4837 }, { "epoch": 1.5, "learning_rate": 8.067613522704541e-06, "loss": 0.1353, "step": 4838 }, { "epoch": 1.5, "learning_rate": 8.067213442688538e-06, "loss": 0.2397, "step": 4839 }, { "epoch": 1.5, "learning_rate": 8.066813362672536e-06, "loss": 0.1124, "step": 4840 }, { "epoch": 1.5, "learning_rate": 8.066413282656533e-06, "loss": 0.1837, "step": 4841 }, { "epoch": 1.5, "learning_rate": 8.06601320264053e-06, "loss": 0.2034, "step": 4842 }, { "epoch": 1.5, "learning_rate": 8.065613122624526e-06, "loss": 0.11, "step": 4843 }, { "epoch": 1.5, "learning_rate": 8.065213042608523e-06, "loss": 0.0984, "step": 4844 }, { "epoch": 1.5, "learning_rate": 8.064812962592518e-06, "loss": 0.1288, "step": 4845 }, { "epoch": 1.5, "learning_rate": 8.064412882576516e-06, "loss": 0.1201, "step": 4846 }, { "epoch": 1.5, "learning_rate": 8.064012802560513e-06, "loss": 0.1062, "step": 4847 }, { "epoch": 1.5, "learning_rate": 8.06361272254451e-06, "loss": 0.1517, "step": 4848 }, { "epoch": 1.51, "learning_rate": 8.063212642528506e-06, "loss": 0.2637, "step": 4849 }, { "epoch": 1.51, "learning_rate": 8.062812562512503e-06, "loss": 0.139, "step": 4850 }, { "epoch": 1.51, "learning_rate": 8.0624124824965e-06, "loss": 0.1482, "step": 4851 }, { "epoch": 1.51, "learning_rate": 8.062012402480496e-06, "loss": 0.141, "step": 4852 }, { "epoch": 1.51, "learning_rate": 8.061612322464493e-06, "loss": 0.1577, "step": 4853 }, { "epoch": 1.51, "learning_rate": 8.061212242448491e-06, "loss": 0.2611, "step": 4854 }, { "epoch": 1.51, "learning_rate": 8.060812162432488e-06, "loss": 0.1674, "step": 4855 }, { "epoch": 1.51, "learning_rate": 8.060412082416485e-06, "loss": 0.0744, "step": 4856 }, { "epoch": 1.51, "learning_rate": 8.06001200240048e-06, "loss": 0.063, "step": 4857 }, { "epoch": 1.51, "learning_rate": 8.059611922384476e-06, "loss": 0.1316, "step": 4858 }, { "epoch": 1.51, "learning_rate": 8.059211842368475e-06, "loss": 0.2146, "step": 4859 }, { "epoch": 1.51, "learning_rate": 8.058811762352472e-06, "loss": 0.1012, "step": 4860 }, { "epoch": 1.51, "learning_rate": 8.058411682336468e-06, "loss": 0.1458, "step": 4861 }, { "epoch": 1.51, "learning_rate": 8.058011602320465e-06, "loss": 0.2118, "step": 4862 }, { "epoch": 1.51, "learning_rate": 8.057611522304462e-06, "loss": 0.1584, "step": 4863 }, { "epoch": 1.51, "learning_rate": 8.057211442288458e-06, "loss": 0.0913, "step": 4864 }, { "epoch": 1.51, "learning_rate": 8.056811362272455e-06, "loss": 0.2571, "step": 4865 }, { "epoch": 1.51, "learning_rate": 8.056411282256452e-06, "loss": 0.1826, "step": 4866 }, { "epoch": 1.51, "learning_rate": 8.056011202240448e-06, "loss": 0.2412, "step": 4867 }, { "epoch": 1.51, "learning_rate": 8.055611122224447e-06, "loss": 0.2382, "step": 4868 }, { "epoch": 1.51, "learning_rate": 8.055211042208443e-06, "loss": 0.0715, "step": 4869 }, { "epoch": 1.51, "learning_rate": 8.054810962192438e-06, "loss": 0.1469, "step": 4870 }, { "epoch": 1.51, "learning_rate": 8.054410882176435e-06, "loss": 0.0975, "step": 4871 }, { "epoch": 1.51, "learning_rate": 8.054010802160432e-06, "loss": 0.0915, "step": 4872 }, { "epoch": 1.51, "learning_rate": 8.05361072214443e-06, "loss": 0.1765, "step": 4873 }, { "epoch": 1.51, "learning_rate": 8.053210642128427e-06, "loss": 0.1575, "step": 4874 }, { "epoch": 1.51, "learning_rate": 8.052810562112423e-06, "loss": 0.2306, "step": 4875 }, { "epoch": 1.51, "learning_rate": 8.05241048209642e-06, "loss": 0.2141, "step": 4876 }, { "epoch": 1.51, "learning_rate": 8.052010402080417e-06, "loss": 0.1634, "step": 4877 }, { "epoch": 1.51, "learning_rate": 8.051610322064414e-06, "loss": 0.0675, "step": 4878 }, { "epoch": 1.51, "learning_rate": 8.05121024204841e-06, "loss": 0.2006, "step": 4879 }, { "epoch": 1.51, "learning_rate": 8.050810162032407e-06, "loss": 0.1491, "step": 4880 }, { "epoch": 1.51, "learning_rate": 8.050410082016404e-06, "loss": 0.1343, "step": 4881 }, { "epoch": 1.52, "learning_rate": 8.050010002000402e-06, "loss": 0.1102, "step": 4882 }, { "epoch": 1.52, "learning_rate": 8.049609921984397e-06, "loss": 0.2238, "step": 4883 }, { "epoch": 1.52, "learning_rate": 8.049209841968394e-06, "loss": 0.1195, "step": 4884 }, { "epoch": 1.52, "learning_rate": 8.04880976195239e-06, "loss": 0.1423, "step": 4885 }, { "epoch": 1.52, "learning_rate": 8.048409681936387e-06, "loss": 0.1505, "step": 4886 }, { "epoch": 1.52, "learning_rate": 8.048009601920385e-06, "loss": 0.0876, "step": 4887 }, { "epoch": 1.52, "learning_rate": 8.047609521904382e-06, "loss": 0.1925, "step": 4888 }, { "epoch": 1.52, "learning_rate": 8.047209441888379e-06, "loss": 0.2635, "step": 4889 }, { "epoch": 1.52, "learning_rate": 8.046809361872375e-06, "loss": 0.151, "step": 4890 }, { "epoch": 1.52, "learning_rate": 8.046409281856372e-06, "loss": 0.1601, "step": 4891 }, { "epoch": 1.52, "learning_rate": 8.046009201840369e-06, "loss": 0.1074, "step": 4892 }, { "epoch": 1.52, "learning_rate": 8.045609121824366e-06, "loss": 0.161, "step": 4893 }, { "epoch": 1.52, "learning_rate": 8.045209041808362e-06, "loss": 0.2261, "step": 4894 }, { "epoch": 1.52, "learning_rate": 8.044808961792359e-06, "loss": 0.1196, "step": 4895 }, { "epoch": 1.52, "learning_rate": 8.044408881776356e-06, "loss": 0.1777, "step": 4896 }, { "epoch": 1.52, "learning_rate": 8.044008801760352e-06, "loss": 0.1744, "step": 4897 }, { "epoch": 1.52, "learning_rate": 8.043608721744349e-06, "loss": 0.1707, "step": 4898 }, { "epoch": 1.52, "learning_rate": 8.043208641728346e-06, "loss": 0.1031, "step": 4899 }, { "epoch": 1.52, "learning_rate": 8.042808561712342e-06, "loss": 0.1147, "step": 4900 }, { "epoch": 1.52, "learning_rate": 8.04240848169634e-06, "loss": 0.1435, "step": 4901 }, { "epoch": 1.52, "learning_rate": 8.042008401680337e-06, "loss": 0.1207, "step": 4902 }, { "epoch": 1.52, "learning_rate": 8.041608321664334e-06, "loss": 0.1396, "step": 4903 }, { "epoch": 1.52, "learning_rate": 8.04120824164833e-06, "loss": 0.1698, "step": 4904 }, { "epoch": 1.52, "learning_rate": 8.040808161632327e-06, "loss": 0.1497, "step": 4905 }, { "epoch": 1.52, "learning_rate": 8.040408081616324e-06, "loss": 0.1518, "step": 4906 }, { "epoch": 1.52, "learning_rate": 8.04000800160032e-06, "loss": 0.1002, "step": 4907 }, { "epoch": 1.52, "learning_rate": 8.039607921584318e-06, "loss": 0.0975, "step": 4908 }, { "epoch": 1.52, "learning_rate": 8.039207841568314e-06, "loss": 0.1218, "step": 4909 }, { "epoch": 1.52, "learning_rate": 8.038807761552311e-06, "loss": 0.2354, "step": 4910 }, { "epoch": 1.52, "learning_rate": 8.038407681536308e-06, "loss": 0.1275, "step": 4911 }, { "epoch": 1.52, "learning_rate": 8.038007601520304e-06, "loss": 0.2888, "step": 4912 }, { "epoch": 1.52, "learning_rate": 8.037607521504301e-06, "loss": 0.2175, "step": 4913 }, { "epoch": 1.53, "learning_rate": 8.037207441488298e-06, "loss": 0.1198, "step": 4914 }, { "epoch": 1.53, "learning_rate": 8.036807361472296e-06, "loss": 0.0869, "step": 4915 }, { "epoch": 1.53, "learning_rate": 8.036407281456293e-06, "loss": 0.3039, "step": 4916 }, { "epoch": 1.53, "learning_rate": 8.03600720144029e-06, "loss": 0.2009, "step": 4917 }, { "epoch": 1.53, "learning_rate": 8.035607121424286e-06, "loss": 0.1365, "step": 4918 }, { "epoch": 1.53, "learning_rate": 8.035207041408283e-06, "loss": 0.134, "step": 4919 }, { "epoch": 1.53, "learning_rate": 8.034806961392278e-06, "loss": 0.1076, "step": 4920 }, { "epoch": 1.53, "learning_rate": 8.034406881376276e-06, "loss": 0.1257, "step": 4921 }, { "epoch": 1.53, "learning_rate": 8.034006801360273e-06, "loss": 0.1509, "step": 4922 }, { "epoch": 1.53, "learning_rate": 8.03360672134427e-06, "loss": 0.0766, "step": 4923 }, { "epoch": 1.53, "learning_rate": 8.033206641328266e-06, "loss": 0.2437, "step": 4924 }, { "epoch": 1.53, "learning_rate": 8.032806561312263e-06, "loss": 0.3028, "step": 4925 }, { "epoch": 1.53, "learning_rate": 8.03240648129626e-06, "loss": 0.1352, "step": 4926 }, { "epoch": 1.53, "learning_rate": 8.032006401280256e-06, "loss": 0.1213, "step": 4927 }, { "epoch": 1.53, "learning_rate": 8.031606321264253e-06, "loss": 0.1789, "step": 4928 }, { "epoch": 1.53, "learning_rate": 8.031206241248251e-06, "loss": 0.1579, "step": 4929 }, { "epoch": 1.53, "learning_rate": 8.030806161232248e-06, "loss": 0.1332, "step": 4930 }, { "epoch": 1.53, "learning_rate": 8.030406081216245e-06, "loss": 0.1372, "step": 4931 }, { "epoch": 1.53, "learning_rate": 8.030006001200241e-06, "loss": 0.2078, "step": 4932 }, { "epoch": 1.53, "learning_rate": 8.029605921184236e-06, "loss": 0.124, "step": 4933 }, { "epoch": 1.53, "learning_rate": 8.029205841168233e-06, "loss": 0.1671, "step": 4934 }, { "epoch": 1.53, "learning_rate": 8.028805761152231e-06, "loss": 0.0872, "step": 4935 }, { "epoch": 1.53, "learning_rate": 8.028405681136228e-06, "loss": 0.2661, "step": 4936 }, { "epoch": 1.53, "learning_rate": 8.028005601120225e-06, "loss": 0.2306, "step": 4937 }, { "epoch": 1.53, "learning_rate": 8.027605521104222e-06, "loss": 0.3098, "step": 4938 }, { "epoch": 1.53, "learning_rate": 8.027205441088218e-06, "loss": 0.0361, "step": 4939 }, { "epoch": 1.53, "learning_rate": 8.026805361072215e-06, "loss": 0.1489, "step": 4940 }, { "epoch": 1.53, "learning_rate": 8.026405281056212e-06, "loss": 0.2207, "step": 4941 }, { "epoch": 1.53, "learning_rate": 8.026005201040208e-06, "loss": 0.1617, "step": 4942 }, { "epoch": 1.53, "learning_rate": 8.025605121024207e-06, "loss": 0.1409, "step": 4943 }, { "epoch": 1.53, "learning_rate": 8.025205041008203e-06, "loss": 0.3319, "step": 4944 }, { "epoch": 1.53, "learning_rate": 8.0248049609922e-06, "loss": 0.0633, "step": 4945 }, { "epoch": 1.54, "learning_rate": 8.024404880976195e-06, "loss": 0.1537, "step": 4946 }, { "epoch": 1.54, "learning_rate": 8.024004800960192e-06, "loss": 0.158, "step": 4947 }, { "epoch": 1.54, "learning_rate": 8.023604720944188e-06, "loss": 0.1502, "step": 4948 }, { "epoch": 1.54, "learning_rate": 8.023204640928187e-06, "loss": 0.1064, "step": 4949 }, { "epoch": 1.54, "learning_rate": 8.022804560912183e-06, "loss": 0.1338, "step": 4950 }, { "epoch": 1.54, "learning_rate": 8.02240448089618e-06, "loss": 0.1741, "step": 4951 }, { "epoch": 1.54, "learning_rate": 8.022004400880177e-06, "loss": 0.1517, "step": 4952 }, { "epoch": 1.54, "learning_rate": 8.021604320864173e-06, "loss": 0.1362, "step": 4953 }, { "epoch": 1.54, "learning_rate": 8.02120424084817e-06, "loss": 0.1252, "step": 4954 }, { "epoch": 1.54, "learning_rate": 8.020804160832167e-06, "loss": 0.1121, "step": 4955 }, { "epoch": 1.54, "learning_rate": 8.020404080816164e-06, "loss": 0.1949, "step": 4956 }, { "epoch": 1.54, "learning_rate": 8.020004000800162e-06, "loss": 0.1304, "step": 4957 }, { "epoch": 1.54, "learning_rate": 8.019603920784157e-06, "loss": 0.129, "step": 4958 }, { "epoch": 1.54, "learning_rate": 8.019203840768154e-06, "loss": 0.0762, "step": 4959 }, { "epoch": 1.54, "learning_rate": 8.01880376075215e-06, "loss": 0.1474, "step": 4960 }, { "epoch": 1.54, "learning_rate": 8.018403680736147e-06, "loss": 0.1229, "step": 4961 }, { "epoch": 1.54, "learning_rate": 8.018003600720144e-06, "loss": 0.2134, "step": 4962 }, { "epoch": 1.54, "learning_rate": 8.017603520704142e-06, "loss": 0.2675, "step": 4963 }, { "epoch": 1.54, "learning_rate": 8.017203440688139e-06, "loss": 0.1128, "step": 4964 }, { "epoch": 1.54, "learning_rate": 8.016803360672135e-06, "loss": 0.0756, "step": 4965 }, { "epoch": 1.54, "learning_rate": 8.016403280656132e-06, "loss": 0.1066, "step": 4966 }, { "epoch": 1.54, "learning_rate": 8.016003200640129e-06, "loss": 0.1356, "step": 4967 }, { "epoch": 1.54, "learning_rate": 8.015603120624125e-06, "loss": 0.2226, "step": 4968 }, { "epoch": 1.54, "learning_rate": 8.015203040608122e-06, "loss": 0.1851, "step": 4969 }, { "epoch": 1.54, "learning_rate": 8.014802960592119e-06, "loss": 0.1348, "step": 4970 }, { "epoch": 1.54, "learning_rate": 8.014402880576116e-06, "loss": 0.2518, "step": 4971 }, { "epoch": 1.54, "learning_rate": 8.014002800560112e-06, "loss": 0.1303, "step": 4972 }, { "epoch": 1.54, "learning_rate": 8.013602720544109e-06, "loss": 0.2699, "step": 4973 }, { "epoch": 1.54, "learning_rate": 8.013202640528106e-06, "loss": 0.1825, "step": 4974 }, { "epoch": 1.54, "learning_rate": 8.012802560512102e-06, "loss": 0.2337, "step": 4975 }, { "epoch": 1.54, "learning_rate": 8.012402480496099e-06, "loss": 0.1195, "step": 4976 }, { "epoch": 1.54, "learning_rate": 8.012002400480097e-06, "loss": 0.1866, "step": 4977 }, { "epoch": 1.55, "learning_rate": 8.011602320464094e-06, "loss": 0.3519, "step": 4978 }, { "epoch": 1.55, "learning_rate": 8.01120224044809e-06, "loss": 0.1579, "step": 4979 }, { "epoch": 1.55, "learning_rate": 8.010802160432087e-06, "loss": 0.0921, "step": 4980 }, { "epoch": 1.55, "learning_rate": 8.010402080416084e-06, "loss": 0.1317, "step": 4981 }, { "epoch": 1.55, "learning_rate": 8.01000200040008e-06, "loss": 0.0765, "step": 4982 }, { "epoch": 1.55, "learning_rate": 8.009601920384077e-06, "loss": 0.2273, "step": 4983 }, { "epoch": 1.55, "learning_rate": 8.009201840368074e-06, "loss": 0.1326, "step": 4984 }, { "epoch": 1.55, "learning_rate": 8.00880176035207e-06, "loss": 0.1202, "step": 4985 }, { "epoch": 1.55, "learning_rate": 8.008401680336068e-06, "loss": 0.2257, "step": 4986 }, { "epoch": 1.55, "learning_rate": 8.008001600320064e-06, "loss": 0.1587, "step": 4987 }, { "epoch": 1.55, "learning_rate": 8.007601520304061e-06, "loss": 0.1165, "step": 4988 }, { "epoch": 1.55, "learning_rate": 8.007201440288058e-06, "loss": 0.1329, "step": 4989 }, { "epoch": 1.55, "learning_rate": 8.006801360272054e-06, "loss": 0.156, "step": 4990 }, { "epoch": 1.55, "learning_rate": 8.006401280256053e-06, "loss": 0.172, "step": 4991 }, { "epoch": 1.55, "learning_rate": 8.00600120024005e-06, "loss": 0.0796, "step": 4992 }, { "epoch": 1.55, "learning_rate": 8.005601120224046e-06, "loss": 0.177, "step": 4993 }, { "epoch": 1.55, "learning_rate": 8.005201040208043e-06, "loss": 0.1502, "step": 4994 }, { "epoch": 1.55, "learning_rate": 8.00480096019204e-06, "loss": 0.1091, "step": 4995 }, { "epoch": 1.55, "learning_rate": 8.004400880176036e-06, "loss": 0.1372, "step": 4996 }, { "epoch": 1.55, "learning_rate": 8.004000800160033e-06, "loss": 0.0961, "step": 4997 }, { "epoch": 1.55, "learning_rate": 8.00360072014403e-06, "loss": 0.1045, "step": 4998 }, { "epoch": 1.55, "learning_rate": 8.003200640128026e-06, "loss": 0.0765, "step": 4999 }, { "epoch": 1.55, "learning_rate": 8.002800560112023e-06, "loss": 0.2863, "step": 5000 }, { "epoch": 1.55, "learning_rate": 8.00240048009602e-06, "loss": 0.1065, "step": 5001 }, { "epoch": 1.55, "learning_rate": 8.002000400080016e-06, "loss": 0.223, "step": 5002 }, { "epoch": 1.55, "learning_rate": 8.001600320064013e-06, "loss": 0.1139, "step": 5003 }, { "epoch": 1.55, "learning_rate": 8.001200240048011e-06, "loss": 0.0967, "step": 5004 }, { "epoch": 1.55, "learning_rate": 8.000800160032008e-06, "loss": 0.1022, "step": 5005 }, { "epoch": 1.55, "learning_rate": 8.000400080016005e-06, "loss": 0.1456, "step": 5006 }, { "epoch": 1.55, "learning_rate": 8.000000000000001e-06, "loss": 0.1585, "step": 5007 }, { "epoch": 1.55, "learning_rate": 7.999599919983998e-06, "loss": 0.0916, "step": 5008 }, { "epoch": 1.55, "learning_rate": 7.999199839967993e-06, "loss": 0.1753, "step": 5009 }, { "epoch": 1.55, "learning_rate": 7.998799759951991e-06, "loss": 0.1614, "step": 5010 }, { "epoch": 1.56, "learning_rate": 7.998399679935988e-06, "loss": 0.1156, "step": 5011 }, { "epoch": 1.56, "learning_rate": 7.997999599919985e-06, "loss": 0.1844, "step": 5012 }, { "epoch": 1.56, "learning_rate": 7.997599519903981e-06, "loss": 0.1019, "step": 5013 }, { "epoch": 1.56, "learning_rate": 7.997199439887978e-06, "loss": 0.0999, "step": 5014 }, { "epoch": 1.56, "learning_rate": 7.996799359871975e-06, "loss": 0.1199, "step": 5015 }, { "epoch": 1.56, "learning_rate": 7.996399279855972e-06, "loss": 0.1209, "step": 5016 }, { "epoch": 1.56, "learning_rate": 7.995999199839968e-06, "loss": 0.1412, "step": 5017 }, { "epoch": 1.56, "learning_rate": 7.995599119823967e-06, "loss": 0.1191, "step": 5018 }, { "epoch": 1.56, "learning_rate": 7.995199039807963e-06, "loss": 0.1042, "step": 5019 }, { "epoch": 1.56, "learning_rate": 7.99479895979196e-06, "loss": 0.1187, "step": 5020 }, { "epoch": 1.56, "learning_rate": 7.994398879775955e-06, "loss": 0.1087, "step": 5021 }, { "epoch": 1.56, "learning_rate": 7.993998799759952e-06, "loss": 0.1137, "step": 5022 }, { "epoch": 1.56, "learning_rate": 7.993598719743948e-06, "loss": 0.1481, "step": 5023 }, { "epoch": 1.56, "learning_rate": 7.993198639727947e-06, "loss": 0.1698, "step": 5024 }, { "epoch": 1.56, "learning_rate": 7.992798559711943e-06, "loss": 0.2823, "step": 5025 }, { "epoch": 1.56, "learning_rate": 7.99239847969594e-06, "loss": 0.2586, "step": 5026 }, { "epoch": 1.56, "learning_rate": 7.991998399679937e-06, "loss": 0.1207, "step": 5027 }, { "epoch": 1.56, "learning_rate": 7.991598319663933e-06, "loss": 0.1378, "step": 5028 }, { "epoch": 1.56, "learning_rate": 7.99119823964793e-06, "loss": 0.1934, "step": 5029 }, { "epoch": 1.56, "learning_rate": 7.990798159631927e-06, "loss": 0.177, "step": 5030 }, { "epoch": 1.56, "learning_rate": 7.990398079615923e-06, "loss": 0.1851, "step": 5031 }, { "epoch": 1.56, "learning_rate": 7.989997999599922e-06, "loss": 0.3083, "step": 5032 }, { "epoch": 1.56, "learning_rate": 7.989597919583919e-06, "loss": 0.1156, "step": 5033 }, { "epoch": 1.56, "learning_rate": 7.989197839567914e-06, "loss": 0.1183, "step": 5034 }, { "epoch": 1.56, "learning_rate": 7.98879775955191e-06, "loss": 0.1856, "step": 5035 }, { "epoch": 1.56, "learning_rate": 7.988397679535907e-06, "loss": 0.1339, "step": 5036 }, { "epoch": 1.56, "learning_rate": 7.987997599519904e-06, "loss": 0.1359, "step": 5037 }, { "epoch": 1.56, "learning_rate": 7.987597519503902e-06, "loss": 0.2064, "step": 5038 }, { "epoch": 1.56, "learning_rate": 7.987197439487899e-06, "loss": 0.1354, "step": 5039 }, { "epoch": 1.56, "learning_rate": 7.986797359471895e-06, "loss": 0.2173, "step": 5040 }, { "epoch": 1.56, "learning_rate": 7.986397279455892e-06, "loss": 0.0795, "step": 5041 }, { "epoch": 1.56, "learning_rate": 7.985997199439889e-06, "loss": 0.1887, "step": 5042 }, { "epoch": 1.57, "learning_rate": 7.985597119423885e-06, "loss": 0.1828, "step": 5043 }, { "epoch": 1.57, "learning_rate": 7.985197039407882e-06, "loss": 0.1989, "step": 5044 }, { "epoch": 1.57, "learning_rate": 7.984796959391879e-06, "loss": 0.1369, "step": 5045 }, { "epoch": 1.57, "learning_rate": 7.984396879375877e-06, "loss": 0.0885, "step": 5046 }, { "epoch": 1.57, "learning_rate": 7.983996799359872e-06, "loss": 0.1347, "step": 5047 }, { "epoch": 1.57, "learning_rate": 7.983596719343869e-06, "loss": 0.0924, "step": 5048 }, { "epoch": 1.57, "learning_rate": 7.983196639327866e-06, "loss": 0.2547, "step": 5049 }, { "epoch": 1.57, "learning_rate": 7.982796559311862e-06, "loss": 0.1094, "step": 5050 }, { "epoch": 1.57, "learning_rate": 7.982396479295859e-06, "loss": 0.1607, "step": 5051 }, { "epoch": 1.57, "learning_rate": 7.981996399279857e-06, "loss": 0.1049, "step": 5052 }, { "epoch": 1.57, "learning_rate": 7.981596319263854e-06, "loss": 0.22, "step": 5053 }, { "epoch": 1.57, "learning_rate": 7.98119623924785e-06, "loss": 0.1646, "step": 5054 }, { "epoch": 1.57, "learning_rate": 7.980796159231847e-06, "loss": 0.0787, "step": 5055 }, { "epoch": 1.57, "learning_rate": 7.980396079215844e-06, "loss": 0.1545, "step": 5056 }, { "epoch": 1.57, "learning_rate": 7.97999599919984e-06, "loss": 0.2657, "step": 5057 }, { "epoch": 1.57, "learning_rate": 7.979595919183837e-06, "loss": 0.0919, "step": 5058 }, { "epoch": 1.57, "learning_rate": 7.979195839167834e-06, "loss": 0.1332, "step": 5059 }, { "epoch": 1.57, "learning_rate": 7.97879575915183e-06, "loss": 0.1135, "step": 5060 }, { "epoch": 1.57, "learning_rate": 7.978395679135827e-06, "loss": 0.1391, "step": 5061 }, { "epoch": 1.57, "learning_rate": 7.977995599119824e-06, "loss": 0.1573, "step": 5062 }, { "epoch": 1.57, "learning_rate": 7.97759551910382e-06, "loss": 0.103, "step": 5063 }, { "epoch": 1.57, "learning_rate": 7.977195439087818e-06, "loss": 0.1821, "step": 5064 }, { "epoch": 1.57, "learning_rate": 7.976795359071814e-06, "loss": 0.1415, "step": 5065 }, { "epoch": 1.57, "learning_rate": 7.976395279055813e-06, "loss": 0.0907, "step": 5066 }, { "epoch": 1.57, "learning_rate": 7.97599519903981e-06, "loss": 0.136, "step": 5067 }, { "epoch": 1.57, "learning_rate": 7.975595119023806e-06, "loss": 0.1494, "step": 5068 }, { "epoch": 1.57, "learning_rate": 7.975195039007803e-06, "loss": 0.1802, "step": 5069 }, { "epoch": 1.57, "learning_rate": 7.9747949589918e-06, "loss": 0.1029, "step": 5070 }, { "epoch": 1.57, "learning_rate": 7.974394878975796e-06, "loss": 0.1207, "step": 5071 }, { "epoch": 1.57, "learning_rate": 7.973994798959793e-06, "loss": 0.2163, "step": 5072 }, { "epoch": 1.57, "learning_rate": 7.97359471894379e-06, "loss": 0.0841, "step": 5073 }, { "epoch": 1.57, "learning_rate": 7.973194638927786e-06, "loss": 0.1877, "step": 5074 }, { "epoch": 1.58, "learning_rate": 7.972794558911783e-06, "loss": 0.2888, "step": 5075 }, { "epoch": 1.58, "learning_rate": 7.97239447889578e-06, "loss": 0.1209, "step": 5076 }, { "epoch": 1.58, "learning_rate": 7.971994398879776e-06, "loss": 0.1653, "step": 5077 }, { "epoch": 1.58, "learning_rate": 7.971594318863773e-06, "loss": 0.2139, "step": 5078 }, { "epoch": 1.58, "learning_rate": 7.97119423884777e-06, "loss": 0.1232, "step": 5079 }, { "epoch": 1.58, "learning_rate": 7.970794158831768e-06, "loss": 0.2585, "step": 5080 }, { "epoch": 1.58, "learning_rate": 7.970394078815765e-06, "loss": 0.249, "step": 5081 }, { "epoch": 1.58, "learning_rate": 7.969993998799761e-06, "loss": 0.2372, "step": 5082 }, { "epoch": 1.58, "learning_rate": 7.969593918783758e-06, "loss": 0.0881, "step": 5083 }, { "epoch": 1.58, "learning_rate": 7.969193838767753e-06, "loss": 0.1676, "step": 5084 }, { "epoch": 1.58, "learning_rate": 7.968793758751751e-06, "loss": 0.1897, "step": 5085 }, { "epoch": 1.58, "learning_rate": 7.968393678735748e-06, "loss": 0.1418, "step": 5086 }, { "epoch": 1.58, "learning_rate": 7.967993598719745e-06, "loss": 0.0968, "step": 5087 }, { "epoch": 1.58, "learning_rate": 7.967593518703741e-06, "loss": 0.1308, "step": 5088 }, { "epoch": 1.58, "learning_rate": 7.967193438687738e-06, "loss": 0.1317, "step": 5089 }, { "epoch": 1.58, "learning_rate": 7.966793358671735e-06, "loss": 0.1155, "step": 5090 }, { "epoch": 1.58, "learning_rate": 7.966393278655731e-06, "loss": 0.1208, "step": 5091 }, { "epoch": 1.58, "learning_rate": 7.965993198639728e-06, "loss": 0.1069, "step": 5092 }, { "epoch": 1.58, "learning_rate": 7.965593118623725e-06, "loss": 0.1263, "step": 5093 }, { "epoch": 1.58, "learning_rate": 7.965193038607723e-06, "loss": 0.1068, "step": 5094 }, { "epoch": 1.58, "learning_rate": 7.96479295859172e-06, "loss": 0.1414, "step": 5095 }, { "epoch": 1.58, "learning_rate": 7.964392878575717e-06, "loss": 0.2051, "step": 5096 }, { "epoch": 1.58, "learning_rate": 7.963992798559712e-06, "loss": 0.1525, "step": 5097 }, { "epoch": 1.58, "learning_rate": 7.963592718543708e-06, "loss": 0.3796, "step": 5098 }, { "epoch": 1.58, "learning_rate": 7.963192638527707e-06, "loss": 0.2679, "step": 5099 }, { "epoch": 1.58, "learning_rate": 7.962792558511703e-06, "loss": 0.168, "step": 5100 }, { "epoch": 1.58, "learning_rate": 7.9623924784957e-06, "loss": 0.1217, "step": 5101 }, { "epoch": 1.58, "learning_rate": 7.961992398479697e-06, "loss": 0.354, "step": 5102 }, { "epoch": 1.58, "learning_rate": 7.961592318463693e-06, "loss": 0.2087, "step": 5103 }, { "epoch": 1.58, "learning_rate": 7.96119223844769e-06, "loss": 0.1922, "step": 5104 }, { "epoch": 1.58, "learning_rate": 7.960792158431687e-06, "loss": 0.1594, "step": 5105 }, { "epoch": 1.58, "learning_rate": 7.960392078415683e-06, "loss": 0.1491, "step": 5106 }, { "epoch": 1.59, "learning_rate": 7.95999199839968e-06, "loss": 0.1444, "step": 5107 }, { "epoch": 1.59, "learning_rate": 7.959591918383679e-06, "loss": 0.1809, "step": 5108 }, { "epoch": 1.59, "learning_rate": 7.959191838367675e-06, "loss": 0.1074, "step": 5109 }, { "epoch": 1.59, "learning_rate": 7.95879175835167e-06, "loss": 0.129, "step": 5110 }, { "epoch": 1.59, "learning_rate": 7.958391678335667e-06, "loss": 0.1311, "step": 5111 }, { "epoch": 1.59, "learning_rate": 7.957991598319664e-06, "loss": 0.1489, "step": 5112 }, { "epoch": 1.59, "learning_rate": 7.957591518303662e-06, "loss": 0.1177, "step": 5113 }, { "epoch": 1.59, "learning_rate": 7.957191438287659e-06, "loss": 0.1908, "step": 5114 }, { "epoch": 1.59, "learning_rate": 7.956791358271655e-06, "loss": 0.102, "step": 5115 }, { "epoch": 1.59, "learning_rate": 7.956391278255652e-06, "loss": 0.1301, "step": 5116 }, { "epoch": 1.59, "learning_rate": 7.955991198239649e-06, "loss": 0.2483, "step": 5117 }, { "epoch": 1.59, "learning_rate": 7.955591118223645e-06, "loss": 0.1955, "step": 5118 }, { "epoch": 1.59, "learning_rate": 7.955191038207642e-06, "loss": 0.1219, "step": 5119 }, { "epoch": 1.59, "learning_rate": 7.954790958191639e-06, "loss": 0.1744, "step": 5120 }, { "epoch": 1.59, "learning_rate": 7.954390878175635e-06, "loss": 0.1228, "step": 5121 }, { "epoch": 1.59, "learning_rate": 7.953990798159632e-06, "loss": 0.1812, "step": 5122 }, { "epoch": 1.59, "learning_rate": 7.953590718143629e-06, "loss": 0.1635, "step": 5123 }, { "epoch": 1.59, "learning_rate": 7.953190638127625e-06, "loss": 0.1483, "step": 5124 }, { "epoch": 1.59, "learning_rate": 7.952790558111622e-06, "loss": 0.1166, "step": 5125 }, { "epoch": 1.59, "learning_rate": 7.952390478095619e-06, "loss": 0.0887, "step": 5126 }, { "epoch": 1.59, "learning_rate": 7.951990398079617e-06, "loss": 0.14, "step": 5127 }, { "epoch": 1.59, "learning_rate": 7.951590318063614e-06, "loss": 0.1558, "step": 5128 }, { "epoch": 1.59, "learning_rate": 7.95119023804761e-06, "loss": 0.2774, "step": 5129 }, { "epoch": 1.59, "learning_rate": 7.950790158031607e-06, "loss": 0.1124, "step": 5130 }, { "epoch": 1.59, "learning_rate": 7.950390078015604e-06, "loss": 0.1192, "step": 5131 }, { "epoch": 1.59, "learning_rate": 7.9499899979996e-06, "loss": 0.1909, "step": 5132 }, { "epoch": 1.59, "learning_rate": 7.949589917983597e-06, "loss": 0.1984, "step": 5133 }, { "epoch": 1.59, "learning_rate": 7.949189837967594e-06, "loss": 0.1663, "step": 5134 }, { "epoch": 1.59, "learning_rate": 7.94878975795159e-06, "loss": 0.2104, "step": 5135 }, { "epoch": 1.59, "learning_rate": 7.948389677935587e-06, "loss": 0.2797, "step": 5136 }, { "epoch": 1.59, "learning_rate": 7.947989597919584e-06, "loss": 0.1201, "step": 5137 }, { "epoch": 1.59, "learning_rate": 7.94758951790358e-06, "loss": 0.1038, "step": 5138 }, { "epoch": 1.6, "learning_rate": 7.947189437887577e-06, "loss": 0.1158, "step": 5139 }, { "epoch": 1.6, "learning_rate": 7.946789357871574e-06, "loss": 0.107, "step": 5140 }, { "epoch": 1.6, "learning_rate": 7.946389277855573e-06, "loss": 0.1227, "step": 5141 }, { "epoch": 1.6, "learning_rate": 7.94598919783957e-06, "loss": 0.1154, "step": 5142 }, { "epoch": 1.6, "learning_rate": 7.945589117823566e-06, "loss": 0.1512, "step": 5143 }, { "epoch": 1.6, "learning_rate": 7.945189037807563e-06, "loss": 0.1149, "step": 5144 }, { "epoch": 1.6, "learning_rate": 7.94478895779156e-06, "loss": 0.1285, "step": 5145 }, { "epoch": 1.6, "learning_rate": 7.944388877775556e-06, "loss": 0.1777, "step": 5146 }, { "epoch": 1.6, "learning_rate": 7.943988797759553e-06, "loss": 0.0933, "step": 5147 }, { "epoch": 1.6, "learning_rate": 7.94358871774355e-06, "loss": 0.1251, "step": 5148 }, { "epoch": 1.6, "learning_rate": 7.943188637727546e-06, "loss": 0.1114, "step": 5149 }, { "epoch": 1.6, "learning_rate": 7.942788557711543e-06, "loss": 0.1796, "step": 5150 }, { "epoch": 1.6, "learning_rate": 7.94238847769554e-06, "loss": 0.1212, "step": 5151 }, { "epoch": 1.6, "learning_rate": 7.941988397679536e-06, "loss": 0.2157, "step": 5152 }, { "epoch": 1.6, "learning_rate": 7.941588317663533e-06, "loss": 0.1498, "step": 5153 }, { "epoch": 1.6, "learning_rate": 7.94118823764753e-06, "loss": 0.1406, "step": 5154 }, { "epoch": 1.6, "learning_rate": 7.940788157631528e-06, "loss": 0.0517, "step": 5155 }, { "epoch": 1.6, "learning_rate": 7.940388077615525e-06, "loss": 0.1918, "step": 5156 }, { "epoch": 1.6, "learning_rate": 7.939987997599521e-06, "loss": 0.1975, "step": 5157 }, { "epoch": 1.6, "learning_rate": 7.939587917583518e-06, "loss": 0.1817, "step": 5158 }, { "epoch": 1.6, "learning_rate": 7.939187837567515e-06, "loss": 0.2913, "step": 5159 }, { "epoch": 1.6, "learning_rate": 7.938787757551511e-06, "loss": 0.1791, "step": 5160 }, { "epoch": 1.6, "learning_rate": 7.938387677535508e-06, "loss": 0.2001, "step": 5161 }, { "epoch": 1.6, "learning_rate": 7.937987597519505e-06, "loss": 0.1062, "step": 5162 }, { "epoch": 1.6, "learning_rate": 7.937587517503501e-06, "loss": 0.1728, "step": 5163 }, { "epoch": 1.6, "learning_rate": 7.937187437487498e-06, "loss": 0.1083, "step": 5164 }, { "epoch": 1.6, "learning_rate": 7.936787357471495e-06, "loss": 0.4234, "step": 5165 }, { "epoch": 1.6, "learning_rate": 7.936387277455491e-06, "loss": 0.1265, "step": 5166 }, { "epoch": 1.6, "learning_rate": 7.935987197439488e-06, "loss": 0.2189, "step": 5167 }, { "epoch": 1.6, "learning_rate": 7.935587117423485e-06, "loss": 0.1677, "step": 5168 }, { "epoch": 1.6, "learning_rate": 7.935187037407483e-06, "loss": 0.1457, "step": 5169 }, { "epoch": 1.6, "learning_rate": 7.93478695739148e-06, "loss": 0.1422, "step": 5170 }, { "epoch": 1.6, "learning_rate": 7.934386877375477e-06, "loss": 0.2673, "step": 5171 }, { "epoch": 1.61, "learning_rate": 7.933986797359473e-06, "loss": 0.1566, "step": 5172 }, { "epoch": 1.61, "learning_rate": 7.933586717343468e-06, "loss": 0.2063, "step": 5173 }, { "epoch": 1.61, "learning_rate": 7.933186637327467e-06, "loss": 0.1326, "step": 5174 }, { "epoch": 1.61, "learning_rate": 7.932786557311463e-06, "loss": 0.1747, "step": 5175 }, { "epoch": 1.61, "learning_rate": 7.93238647729546e-06, "loss": 0.1605, "step": 5176 }, { "epoch": 1.61, "learning_rate": 7.931986397279457e-06, "loss": 0.1616, "step": 5177 }, { "epoch": 1.61, "learning_rate": 7.931586317263453e-06, "loss": 0.1947, "step": 5178 }, { "epoch": 1.61, "learning_rate": 7.93118623724745e-06, "loss": 0.1277, "step": 5179 }, { "epoch": 1.61, "learning_rate": 7.930786157231447e-06, "loss": 0.0991, "step": 5180 }, { "epoch": 1.61, "learning_rate": 7.930386077215443e-06, "loss": 0.1203, "step": 5181 }, { "epoch": 1.61, "learning_rate": 7.92998599719944e-06, "loss": 0.1472, "step": 5182 }, { "epoch": 1.61, "learning_rate": 7.929585917183438e-06, "loss": 0.1581, "step": 5183 }, { "epoch": 1.61, "learning_rate": 7.929185837167435e-06, "loss": 0.1227, "step": 5184 }, { "epoch": 1.61, "learning_rate": 7.92878575715143e-06, "loss": 0.124, "step": 5185 }, { "epoch": 1.61, "learning_rate": 7.928385677135427e-06, "loss": 0.168, "step": 5186 }, { "epoch": 1.61, "learning_rate": 7.927985597119423e-06, "loss": 0.0935, "step": 5187 }, { "epoch": 1.61, "learning_rate": 7.927585517103422e-06, "loss": 0.1182, "step": 5188 }, { "epoch": 1.61, "learning_rate": 7.927185437087419e-06, "loss": 0.0746, "step": 5189 }, { "epoch": 1.61, "learning_rate": 7.926785357071415e-06, "loss": 0.0966, "step": 5190 }, { "epoch": 1.61, "learning_rate": 7.926385277055412e-06, "loss": 0.1131, "step": 5191 }, { "epoch": 1.61, "learning_rate": 7.925985197039409e-06, "loss": 0.1252, "step": 5192 }, { "epoch": 1.61, "learning_rate": 7.925585117023405e-06, "loss": 0.2616, "step": 5193 }, { "epoch": 1.61, "learning_rate": 7.925185037007402e-06, "loss": 0.1212, "step": 5194 }, { "epoch": 1.61, "learning_rate": 7.924784956991399e-06, "loss": 0.1033, "step": 5195 }, { "epoch": 1.61, "learning_rate": 7.924384876975395e-06, "loss": 0.1057, "step": 5196 }, { "epoch": 1.61, "learning_rate": 7.923984796959394e-06, "loss": 0.2041, "step": 5197 }, { "epoch": 1.61, "learning_rate": 7.923584716943389e-06, "loss": 0.1735, "step": 5198 }, { "epoch": 1.61, "learning_rate": 7.923184636927385e-06, "loss": 0.1869, "step": 5199 }, { "epoch": 1.61, "learning_rate": 7.922784556911382e-06, "loss": 0.1475, "step": 5200 }, { "epoch": 1.61, "learning_rate": 7.922384476895379e-06, "loss": 0.1957, "step": 5201 }, { "epoch": 1.61, "learning_rate": 7.921984396879377e-06, "loss": 0.2475, "step": 5202 }, { "epoch": 1.61, "learning_rate": 7.921584316863374e-06, "loss": 0.104, "step": 5203 }, { "epoch": 1.62, "learning_rate": 7.92118423684737e-06, "loss": 0.1672, "step": 5204 }, { "epoch": 1.62, "learning_rate": 7.920784156831367e-06, "loss": 0.1737, "step": 5205 }, { "epoch": 1.62, "learning_rate": 7.920384076815364e-06, "loss": 0.1269, "step": 5206 }, { "epoch": 1.62, "learning_rate": 7.91998399679936e-06, "loss": 0.088, "step": 5207 }, { "epoch": 1.62, "learning_rate": 7.919583916783357e-06, "loss": 0.2942, "step": 5208 }, { "epoch": 1.62, "learning_rate": 7.919183836767354e-06, "loss": 0.0989, "step": 5209 }, { "epoch": 1.62, "learning_rate": 7.91878375675135e-06, "loss": 0.1643, "step": 5210 }, { "epoch": 1.62, "learning_rate": 7.918383676735347e-06, "loss": 0.1316, "step": 5211 }, { "epoch": 1.62, "learning_rate": 7.917983596719344e-06, "loss": 0.117, "step": 5212 }, { "epoch": 1.62, "learning_rate": 7.91758351670334e-06, "loss": 0.1668, "step": 5213 }, { "epoch": 1.62, "learning_rate": 7.917183436687337e-06, "loss": 0.1542, "step": 5214 }, { "epoch": 1.62, "learning_rate": 7.916783356671334e-06, "loss": 0.0611, "step": 5215 }, { "epoch": 1.62, "learning_rate": 7.916383276655332e-06, "loss": 0.1757, "step": 5216 }, { "epoch": 1.62, "learning_rate": 7.91598319663933e-06, "loss": 0.0947, "step": 5217 }, { "epoch": 1.62, "learning_rate": 7.915583116623326e-06, "loss": 0.1201, "step": 5218 }, { "epoch": 1.62, "learning_rate": 7.915183036607323e-06, "loss": 0.128, "step": 5219 }, { "epoch": 1.62, "learning_rate": 7.91478295659132e-06, "loss": 0.0787, "step": 5220 }, { "epoch": 1.62, "learning_rate": 7.914382876575316e-06, "loss": 0.1376, "step": 5221 }, { "epoch": 1.62, "learning_rate": 7.913982796559313e-06, "loss": 0.1437, "step": 5222 }, { "epoch": 1.62, "learning_rate": 7.91358271654331e-06, "loss": 0.1939, "step": 5223 }, { "epoch": 1.62, "learning_rate": 7.913182636527306e-06, "loss": 0.1214, "step": 5224 }, { "epoch": 1.62, "learning_rate": 7.912782556511303e-06, "loss": 0.2201, "step": 5225 }, { "epoch": 1.62, "learning_rate": 7.9123824764953e-06, "loss": 0.1609, "step": 5226 }, { "epoch": 1.62, "learning_rate": 7.911982396479296e-06, "loss": 0.1804, "step": 5227 }, { "epoch": 1.62, "learning_rate": 7.911582316463293e-06, "loss": 0.1299, "step": 5228 }, { "epoch": 1.62, "learning_rate": 7.91118223644729e-06, "loss": 0.2701, "step": 5229 }, { "epoch": 1.62, "learning_rate": 7.910782156431288e-06, "loss": 0.2626, "step": 5230 }, { "epoch": 1.62, "learning_rate": 7.910382076415284e-06, "loss": 0.2023, "step": 5231 }, { "epoch": 1.62, "learning_rate": 7.909981996399281e-06, "loss": 0.0941, "step": 5232 }, { "epoch": 1.62, "learning_rate": 7.909581916383278e-06, "loss": 0.1422, "step": 5233 }, { "epoch": 1.62, "learning_rate": 7.909181836367275e-06, "loss": 0.0999, "step": 5234 }, { "epoch": 1.62, "learning_rate": 7.908781756351271e-06, "loss": 0.17, "step": 5235 }, { "epoch": 1.63, "learning_rate": 7.908381676335268e-06, "loss": 0.2244, "step": 5236 }, { "epoch": 1.63, "learning_rate": 7.907981596319265e-06, "loss": 0.2414, "step": 5237 }, { "epoch": 1.63, "learning_rate": 7.907581516303261e-06, "loss": 0.2096, "step": 5238 }, { "epoch": 1.63, "learning_rate": 7.907181436287258e-06, "loss": 0.162, "step": 5239 }, { "epoch": 1.63, "learning_rate": 7.906781356271255e-06, "loss": 0.1293, "step": 5240 }, { "epoch": 1.63, "learning_rate": 7.906381276255251e-06, "loss": 0.0713, "step": 5241 }, { "epoch": 1.63, "learning_rate": 7.905981196239248e-06, "loss": 0.2058, "step": 5242 }, { "epoch": 1.63, "learning_rate": 7.905581116223245e-06, "loss": 0.1138, "step": 5243 }, { "epoch": 1.63, "learning_rate": 7.905181036207243e-06, "loss": 0.1272, "step": 5244 }, { "epoch": 1.63, "learning_rate": 7.90478095619124e-06, "loss": 0.1016, "step": 5245 }, { "epoch": 1.63, "learning_rate": 7.904380876175236e-06, "loss": 0.1262, "step": 5246 }, { "epoch": 1.63, "learning_rate": 7.903980796159233e-06, "loss": 0.106, "step": 5247 }, { "epoch": 1.63, "learning_rate": 7.903580716143228e-06, "loss": 0.1238, "step": 5248 }, { "epoch": 1.63, "learning_rate": 7.903180636127225e-06, "loss": 0.1894, "step": 5249 }, { "epoch": 1.63, "learning_rate": 7.902780556111223e-06, "loss": 0.1243, "step": 5250 }, { "epoch": 1.63, "learning_rate": 7.90238047609522e-06, "loss": 0.2794, "step": 5251 }, { "epoch": 1.63, "learning_rate": 7.901980396079217e-06, "loss": 0.0897, "step": 5252 }, { "epoch": 1.63, "learning_rate": 7.901580316063213e-06, "loss": 0.2041, "step": 5253 }, { "epoch": 1.63, "learning_rate": 7.90118023604721e-06, "loss": 0.0928, "step": 5254 }, { "epoch": 1.63, "learning_rate": 7.900780156031207e-06, "loss": 0.1081, "step": 5255 }, { "epoch": 1.63, "learning_rate": 7.900380076015203e-06, "loss": 0.1211, "step": 5256 }, { "epoch": 1.63, "learning_rate": 7.8999799959992e-06, "loss": 0.2079, "step": 5257 }, { "epoch": 1.63, "learning_rate": 7.899579915983198e-06, "loss": 0.1754, "step": 5258 }, { "epoch": 1.63, "learning_rate": 7.899179835967195e-06, "loss": 0.1844, "step": 5259 }, { "epoch": 1.63, "learning_rate": 7.898779755951192e-06, "loss": 0.0412, "step": 5260 }, { "epoch": 1.63, "learning_rate": 7.898379675935187e-06, "loss": 0.0935, "step": 5261 }, { "epoch": 1.63, "learning_rate": 7.897979595919183e-06, "loss": 0.1854, "step": 5262 }, { "epoch": 1.63, "learning_rate": 7.89757951590318e-06, "loss": 0.1291, "step": 5263 }, { "epoch": 1.63, "learning_rate": 7.897179435887179e-06, "loss": 0.1563, "step": 5264 }, { "epoch": 1.63, "learning_rate": 7.896779355871175e-06, "loss": 0.1568, "step": 5265 }, { "epoch": 1.63, "learning_rate": 7.896379275855172e-06, "loss": 0.3326, "step": 5266 }, { "epoch": 1.63, "learning_rate": 7.895979195839169e-06, "loss": 0.1045, "step": 5267 }, { "epoch": 1.64, "learning_rate": 7.895579115823165e-06, "loss": 0.1701, "step": 5268 }, { "epoch": 1.64, "learning_rate": 7.895179035807162e-06, "loss": 0.0987, "step": 5269 }, { "epoch": 1.64, "learning_rate": 7.894778955791159e-06, "loss": 0.1007, "step": 5270 }, { "epoch": 1.64, "learning_rate": 7.894378875775155e-06, "loss": 0.1548, "step": 5271 }, { "epoch": 1.64, "learning_rate": 7.893978795759154e-06, "loss": 0.2543, "step": 5272 }, { "epoch": 1.64, "learning_rate": 7.89357871574315e-06, "loss": 0.2375, "step": 5273 }, { "epoch": 1.64, "learning_rate": 7.893178635727145e-06, "loss": 0.1496, "step": 5274 }, { "epoch": 1.64, "learning_rate": 7.892778555711142e-06, "loss": 0.1505, "step": 5275 }, { "epoch": 1.64, "learning_rate": 7.892378475695139e-06, "loss": 0.1183, "step": 5276 }, { "epoch": 1.64, "learning_rate": 7.891978395679135e-06, "loss": 0.2457, "step": 5277 }, { "epoch": 1.64, "learning_rate": 7.891578315663134e-06, "loss": 0.1625, "step": 5278 }, { "epoch": 1.64, "learning_rate": 7.89117823564713e-06, "loss": 0.1354, "step": 5279 }, { "epoch": 1.64, "learning_rate": 7.890778155631127e-06, "loss": 0.1326, "step": 5280 }, { "epoch": 1.64, "learning_rate": 7.890378075615124e-06, "loss": 0.1114, "step": 5281 }, { "epoch": 1.64, "learning_rate": 7.88997799559912e-06, "loss": 0.0731, "step": 5282 }, { "epoch": 1.64, "learning_rate": 7.889577915583117e-06, "loss": 0.1492, "step": 5283 }, { "epoch": 1.64, "learning_rate": 7.889177835567114e-06, "loss": 0.1694, "step": 5284 }, { "epoch": 1.64, "learning_rate": 7.88877775555111e-06, "loss": 0.1119, "step": 5285 }, { "epoch": 1.64, "learning_rate": 7.888377675535107e-06, "loss": 0.139, "step": 5286 }, { "epoch": 1.64, "learning_rate": 7.887977595519104e-06, "loss": 0.1718, "step": 5287 }, { "epoch": 1.64, "learning_rate": 7.8875775155031e-06, "loss": 0.1063, "step": 5288 }, { "epoch": 1.64, "learning_rate": 7.887177435487097e-06, "loss": 0.0719, "step": 5289 }, { "epoch": 1.64, "learning_rate": 7.886777355471094e-06, "loss": 0.1364, "step": 5290 }, { "epoch": 1.64, "learning_rate": 7.88637727545509e-06, "loss": 0.1317, "step": 5291 }, { "epoch": 1.64, "learning_rate": 7.885977195439089e-06, "loss": 0.1935, "step": 5292 }, { "epoch": 1.64, "learning_rate": 7.885577115423086e-06, "loss": 0.097, "step": 5293 }, { "epoch": 1.64, "learning_rate": 7.885177035407082e-06, "loss": 0.1641, "step": 5294 }, { "epoch": 1.64, "learning_rate": 7.88477695539108e-06, "loss": 0.2261, "step": 5295 }, { "epoch": 1.64, "learning_rate": 7.884376875375076e-06, "loss": 0.2441, "step": 5296 }, { "epoch": 1.64, "learning_rate": 7.883976795359073e-06, "loss": 0.272, "step": 5297 }, { "epoch": 1.64, "learning_rate": 7.88357671534307e-06, "loss": 0.1018, "step": 5298 }, { "epoch": 1.64, "learning_rate": 7.883176635327066e-06, "loss": 0.1819, "step": 5299 }, { "epoch": 1.65, "learning_rate": 7.882776555311063e-06, "loss": 0.1306, "step": 5300 }, { "epoch": 1.65, "learning_rate": 7.88237647529506e-06, "loss": 0.186, "step": 5301 }, { "epoch": 1.65, "learning_rate": 7.881976395279056e-06, "loss": 0.1457, "step": 5302 }, { "epoch": 1.65, "learning_rate": 7.881576315263053e-06, "loss": 0.1668, "step": 5303 }, { "epoch": 1.65, "learning_rate": 7.88117623524705e-06, "loss": 0.2317, "step": 5304 }, { "epoch": 1.65, "learning_rate": 7.880776155231048e-06, "loss": 0.1607, "step": 5305 }, { "epoch": 1.65, "learning_rate": 7.880376075215044e-06, "loss": 0.2027, "step": 5306 }, { "epoch": 1.65, "learning_rate": 7.879975995199041e-06, "loss": 0.2372, "step": 5307 }, { "epoch": 1.65, "learning_rate": 7.879575915183038e-06, "loss": 0.0767, "step": 5308 }, { "epoch": 1.65, "learning_rate": 7.879175835167034e-06, "loss": 0.1333, "step": 5309 }, { "epoch": 1.65, "learning_rate": 7.878775755151031e-06, "loss": 0.1165, "step": 5310 }, { "epoch": 1.65, "learning_rate": 7.878375675135028e-06, "loss": 0.1736, "step": 5311 }, { "epoch": 1.65, "learning_rate": 7.877975595119025e-06, "loss": 0.2962, "step": 5312 }, { "epoch": 1.65, "learning_rate": 7.877575515103021e-06, "loss": 0.1051, "step": 5313 }, { "epoch": 1.65, "learning_rate": 7.877175435087018e-06, "loss": 0.1092, "step": 5314 }, { "epoch": 1.65, "learning_rate": 7.876775355071015e-06, "loss": 0.1855, "step": 5315 }, { "epoch": 1.65, "learning_rate": 7.876375275055011e-06, "loss": 0.0987, "step": 5316 }, { "epoch": 1.65, "learning_rate": 7.875975195039008e-06, "loss": 0.1125, "step": 5317 }, { "epoch": 1.65, "learning_rate": 7.875575115023005e-06, "loss": 0.0662, "step": 5318 }, { "epoch": 1.65, "learning_rate": 7.875175035007003e-06, "loss": 0.1267, "step": 5319 }, { "epoch": 1.65, "learning_rate": 7.874774954991e-06, "loss": 0.0666, "step": 5320 }, { "epoch": 1.65, "learning_rate": 7.874374874974996e-06, "loss": 0.1809, "step": 5321 }, { "epoch": 1.65, "learning_rate": 7.873974794958993e-06, "loss": 0.0976, "step": 5322 }, { "epoch": 1.65, "learning_rate": 7.87357471494299e-06, "loss": 0.1675, "step": 5323 }, { "epoch": 1.65, "learning_rate": 7.873174634926985e-06, "loss": 0.1605, "step": 5324 }, { "epoch": 1.65, "learning_rate": 7.872774554910983e-06, "loss": 0.1702, "step": 5325 }, { "epoch": 1.65, "learning_rate": 7.87237447489498e-06, "loss": 0.3082, "step": 5326 }, { "epoch": 1.65, "learning_rate": 7.871974394878977e-06, "loss": 0.1237, "step": 5327 }, { "epoch": 1.65, "learning_rate": 7.871574314862973e-06, "loss": 0.1267, "step": 5328 }, { "epoch": 1.65, "learning_rate": 7.87117423484697e-06, "loss": 0.1324, "step": 5329 }, { "epoch": 1.65, "learning_rate": 7.870774154830967e-06, "loss": 0.2415, "step": 5330 }, { "epoch": 1.65, "learning_rate": 7.870374074814963e-06, "loss": 0.1796, "step": 5331 }, { "epoch": 1.65, "learning_rate": 7.86997399479896e-06, "loss": 0.1014, "step": 5332 }, { "epoch": 1.66, "learning_rate": 7.869573914782958e-06, "loss": 0.0947, "step": 5333 }, { "epoch": 1.66, "learning_rate": 7.869173834766955e-06, "loss": 0.1707, "step": 5334 }, { "epoch": 1.66, "learning_rate": 7.868773754750952e-06, "loss": 0.272, "step": 5335 }, { "epoch": 1.66, "learning_rate": 7.868373674734948e-06, "loss": 0.1359, "step": 5336 }, { "epoch": 1.66, "learning_rate": 7.867973594718943e-06, "loss": 0.1137, "step": 5337 }, { "epoch": 1.66, "learning_rate": 7.86757351470294e-06, "loss": 0.225, "step": 5338 }, { "epoch": 1.66, "learning_rate": 7.867173434686938e-06, "loss": 0.1632, "step": 5339 }, { "epoch": 1.66, "learning_rate": 7.866773354670935e-06, "loss": 0.1222, "step": 5340 }, { "epoch": 1.66, "learning_rate": 7.866373274654932e-06, "loss": 0.1454, "step": 5341 }, { "epoch": 1.66, "learning_rate": 7.865973194638929e-06, "loss": 0.0773, "step": 5342 }, { "epoch": 1.66, "learning_rate": 7.865573114622925e-06, "loss": 0.07, "step": 5343 }, { "epoch": 1.66, "learning_rate": 7.865173034606922e-06, "loss": 0.2155, "step": 5344 }, { "epoch": 1.66, "learning_rate": 7.864772954590919e-06, "loss": 0.3048, "step": 5345 }, { "epoch": 1.66, "learning_rate": 7.864372874574915e-06, "loss": 0.2349, "step": 5346 }, { "epoch": 1.66, "learning_rate": 7.863972794558914e-06, "loss": 0.2171, "step": 5347 }, { "epoch": 1.66, "learning_rate": 7.86357271454291e-06, "loss": 0.146, "step": 5348 }, { "epoch": 1.66, "learning_rate": 7.863172634526905e-06, "loss": 0.1259, "step": 5349 }, { "epoch": 1.66, "learning_rate": 7.862772554510902e-06, "loss": 0.0896, "step": 5350 }, { "epoch": 1.66, "learning_rate": 7.862372474494899e-06, "loss": 0.1943, "step": 5351 }, { "epoch": 1.66, "learning_rate": 7.861972394478895e-06, "loss": 0.1345, "step": 5352 }, { "epoch": 1.66, "learning_rate": 7.861572314462894e-06, "loss": 0.2931, "step": 5353 }, { "epoch": 1.66, "learning_rate": 7.86117223444689e-06, "loss": 0.2226, "step": 5354 }, { "epoch": 1.66, "learning_rate": 7.860772154430887e-06, "loss": 0.139, "step": 5355 }, { "epoch": 1.66, "learning_rate": 7.860372074414884e-06, "loss": 0.124, "step": 5356 }, { "epoch": 1.66, "learning_rate": 7.85997199439888e-06, "loss": 0.1427, "step": 5357 }, { "epoch": 1.66, "learning_rate": 7.859571914382877e-06, "loss": 0.1115, "step": 5358 }, { "epoch": 1.66, "learning_rate": 7.859171834366874e-06, "loss": 0.1131, "step": 5359 }, { "epoch": 1.66, "learning_rate": 7.85877175435087e-06, "loss": 0.2112, "step": 5360 }, { "epoch": 1.66, "learning_rate": 7.858371674334869e-06, "loss": 0.1587, "step": 5361 }, { "epoch": 1.66, "learning_rate": 7.857971594318864e-06, "loss": 0.1152, "step": 5362 }, { "epoch": 1.66, "learning_rate": 7.85757151430286e-06, "loss": 0.1731, "step": 5363 }, { "epoch": 1.66, "learning_rate": 7.857171434286857e-06, "loss": 0.1197, "step": 5364 }, { "epoch": 1.67, "learning_rate": 7.856771354270854e-06, "loss": 0.1419, "step": 5365 }, { "epoch": 1.67, "learning_rate": 7.85637127425485e-06, "loss": 0.1027, "step": 5366 }, { "epoch": 1.67, "learning_rate": 7.855971194238849e-06, "loss": 0.1911, "step": 5367 }, { "epoch": 1.67, "learning_rate": 7.855571114222846e-06, "loss": 0.2991, "step": 5368 }, { "epoch": 1.67, "learning_rate": 7.855171034206842e-06, "loss": 0.1029, "step": 5369 }, { "epoch": 1.67, "learning_rate": 7.854770954190839e-06, "loss": 0.2281, "step": 5370 }, { "epoch": 1.67, "learning_rate": 7.854370874174836e-06, "loss": 0.1137, "step": 5371 }, { "epoch": 1.67, "learning_rate": 7.853970794158832e-06, "loss": 0.1449, "step": 5372 }, { "epoch": 1.67, "learning_rate": 7.85357071414283e-06, "loss": 0.09, "step": 5373 }, { "epoch": 1.67, "learning_rate": 7.853170634126826e-06, "loss": 0.086, "step": 5374 }, { "epoch": 1.67, "learning_rate": 7.852770554110823e-06, "loss": 0.0896, "step": 5375 }, { "epoch": 1.67, "learning_rate": 7.85237047409482e-06, "loss": 0.15, "step": 5376 }, { "epoch": 1.67, "learning_rate": 7.851970394078816e-06, "loss": 0.1495, "step": 5377 }, { "epoch": 1.67, "learning_rate": 7.851570314062813e-06, "loss": 0.1716, "step": 5378 }, { "epoch": 1.67, "learning_rate": 7.85117023404681e-06, "loss": 0.2176, "step": 5379 }, { "epoch": 1.67, "learning_rate": 7.850770154030806e-06, "loss": 0.2694, "step": 5380 }, { "epoch": 1.67, "learning_rate": 7.850370074014804e-06, "loss": 0.1663, "step": 5381 }, { "epoch": 1.67, "learning_rate": 7.849969993998801e-06, "loss": 0.2143, "step": 5382 }, { "epoch": 1.67, "learning_rate": 7.849569913982798e-06, "loss": 0.08, "step": 5383 }, { "epoch": 1.67, "learning_rate": 7.849169833966794e-06, "loss": 0.1461, "step": 5384 }, { "epoch": 1.67, "learning_rate": 7.848769753950791e-06, "loss": 0.1188, "step": 5385 }, { "epoch": 1.67, "learning_rate": 7.848369673934788e-06, "loss": 0.3302, "step": 5386 }, { "epoch": 1.67, "learning_rate": 7.847969593918784e-06, "loss": 0.1027, "step": 5387 }, { "epoch": 1.67, "learning_rate": 7.847569513902781e-06, "loss": 0.0738, "step": 5388 }, { "epoch": 1.67, "learning_rate": 7.847169433886778e-06, "loss": 0.0947, "step": 5389 }, { "epoch": 1.67, "learning_rate": 7.846769353870775e-06, "loss": 0.1328, "step": 5390 }, { "epoch": 1.67, "learning_rate": 7.846369273854771e-06, "loss": 0.1387, "step": 5391 }, { "epoch": 1.67, "learning_rate": 7.845969193838768e-06, "loss": 0.1127, "step": 5392 }, { "epoch": 1.67, "learning_rate": 7.845569113822765e-06, "loss": 0.0647, "step": 5393 }, { "epoch": 1.67, "learning_rate": 7.845169033806761e-06, "loss": 0.1447, "step": 5394 }, { "epoch": 1.67, "learning_rate": 7.84476895379076e-06, "loss": 0.1844, "step": 5395 }, { "epoch": 1.67, "learning_rate": 7.844368873774756e-06, "loss": 0.1903, "step": 5396 }, { "epoch": 1.68, "learning_rate": 7.843968793758753e-06, "loss": 0.2414, "step": 5397 }, { "epoch": 1.68, "learning_rate": 7.84356871374275e-06, "loss": 0.1229, "step": 5398 }, { "epoch": 1.68, "learning_rate": 7.843168633726746e-06, "loss": 0.1791, "step": 5399 }, { "epoch": 1.68, "learning_rate": 7.842768553710743e-06, "loss": 0.1883, "step": 5400 }, { "epoch": 1.68, "learning_rate": 7.84236847369474e-06, "loss": 0.1852, "step": 5401 }, { "epoch": 1.68, "learning_rate": 7.841968393678736e-06, "loss": 0.1693, "step": 5402 }, { "epoch": 1.68, "learning_rate": 7.841568313662733e-06, "loss": 0.1205, "step": 5403 }, { "epoch": 1.68, "learning_rate": 7.84116823364673e-06, "loss": 0.2955, "step": 5404 }, { "epoch": 1.68, "learning_rate": 7.840768153630727e-06, "loss": 0.1229, "step": 5405 }, { "epoch": 1.68, "learning_rate": 7.840368073614723e-06, "loss": 0.1682, "step": 5406 }, { "epoch": 1.68, "learning_rate": 7.83996799359872e-06, "loss": 0.1512, "step": 5407 }, { "epoch": 1.68, "learning_rate": 7.839567913582717e-06, "loss": 0.1929, "step": 5408 }, { "epoch": 1.68, "learning_rate": 7.839167833566715e-06, "loss": 0.1291, "step": 5409 }, { "epoch": 1.68, "learning_rate": 7.838767753550712e-06, "loss": 0.1788, "step": 5410 }, { "epoch": 1.68, "learning_rate": 7.838367673534708e-06, "loss": 0.189, "step": 5411 }, { "epoch": 1.68, "learning_rate": 7.837967593518703e-06, "loss": 0.2115, "step": 5412 }, { "epoch": 1.68, "learning_rate": 7.8375675135027e-06, "loss": 0.2363, "step": 5413 }, { "epoch": 1.68, "learning_rate": 7.837167433486698e-06, "loss": 0.276, "step": 5414 }, { "epoch": 1.68, "learning_rate": 7.836767353470695e-06, "loss": 0.1159, "step": 5415 }, { "epoch": 1.68, "learning_rate": 7.836367273454692e-06, "loss": 0.1846, "step": 5416 }, { "epoch": 1.68, "learning_rate": 7.835967193438688e-06, "loss": 0.0783, "step": 5417 }, { "epoch": 1.68, "learning_rate": 7.835567113422685e-06, "loss": 0.2134, "step": 5418 }, { "epoch": 1.68, "learning_rate": 7.835167033406682e-06, "loss": 0.1366, "step": 5419 }, { "epoch": 1.68, "learning_rate": 7.834766953390679e-06, "loss": 0.1323, "step": 5420 }, { "epoch": 1.68, "learning_rate": 7.834366873374675e-06, "loss": 0.1383, "step": 5421 }, { "epoch": 1.68, "learning_rate": 7.833966793358672e-06, "loss": 0.325, "step": 5422 }, { "epoch": 1.68, "learning_rate": 7.83356671334267e-06, "loss": 0.1396, "step": 5423 }, { "epoch": 1.68, "learning_rate": 7.833166633326667e-06, "loss": 0.1825, "step": 5424 }, { "epoch": 1.68, "learning_rate": 7.832766553310662e-06, "loss": 0.1395, "step": 5425 }, { "epoch": 1.68, "learning_rate": 7.832366473294659e-06, "loss": 0.2995, "step": 5426 }, { "epoch": 1.68, "learning_rate": 7.831966393278655e-06, "loss": 0.2212, "step": 5427 }, { "epoch": 1.68, "learning_rate": 7.831566313262654e-06, "loss": 0.0982, "step": 5428 }, { "epoch": 1.69, "learning_rate": 7.83116623324665e-06, "loss": 0.2006, "step": 5429 }, { "epoch": 1.69, "learning_rate": 7.830766153230647e-06, "loss": 0.1799, "step": 5430 }, { "epoch": 1.69, "learning_rate": 7.830366073214644e-06, "loss": 0.0973, "step": 5431 }, { "epoch": 1.69, "learning_rate": 7.82996599319864e-06, "loss": 0.1977, "step": 5432 }, { "epoch": 1.69, "learning_rate": 7.829565913182637e-06, "loss": 0.2049, "step": 5433 }, { "epoch": 1.69, "learning_rate": 7.829165833166634e-06, "loss": 0.2439, "step": 5434 }, { "epoch": 1.69, "learning_rate": 7.82876575315063e-06, "loss": 0.1125, "step": 5435 }, { "epoch": 1.69, "learning_rate": 7.828365673134629e-06, "loss": 0.2483, "step": 5436 }, { "epoch": 1.69, "learning_rate": 7.827965593118626e-06, "loss": 0.1206, "step": 5437 }, { "epoch": 1.69, "learning_rate": 7.82756551310262e-06, "loss": 0.0768, "step": 5438 }, { "epoch": 1.69, "learning_rate": 7.827165433086617e-06, "loss": 0.2526, "step": 5439 }, { "epoch": 1.69, "learning_rate": 7.826765353070614e-06, "loss": 0.1276, "step": 5440 }, { "epoch": 1.69, "learning_rate": 7.82636527305461e-06, "loss": 0.1383, "step": 5441 }, { "epoch": 1.69, "learning_rate": 7.825965193038609e-06, "loss": 0.1443, "step": 5442 }, { "epoch": 1.69, "learning_rate": 7.825565113022606e-06, "loss": 0.153, "step": 5443 }, { "epoch": 1.69, "learning_rate": 7.825165033006602e-06, "loss": 0.2645, "step": 5444 }, { "epoch": 1.69, "learning_rate": 7.824764952990599e-06, "loss": 0.1807, "step": 5445 }, { "epoch": 1.69, "learning_rate": 7.824364872974596e-06, "loss": 0.1054, "step": 5446 }, { "epoch": 1.69, "learning_rate": 7.823964792958592e-06, "loss": 0.1349, "step": 5447 }, { "epoch": 1.69, "learning_rate": 7.823564712942589e-06, "loss": 0.1293, "step": 5448 }, { "epoch": 1.69, "learning_rate": 7.823164632926586e-06, "loss": 0.1131, "step": 5449 }, { "epoch": 1.69, "learning_rate": 7.822764552910582e-06, "loss": 0.1804, "step": 5450 }, { "epoch": 1.69, "learning_rate": 7.82236447289458e-06, "loss": 0.1117, "step": 5451 }, { "epoch": 1.69, "learning_rate": 7.821964392878576e-06, "loss": 0.118, "step": 5452 }, { "epoch": 1.69, "learning_rate": 7.821564312862573e-06, "loss": 0.0909, "step": 5453 }, { "epoch": 1.69, "learning_rate": 7.82116423284657e-06, "loss": 0.18, "step": 5454 }, { "epoch": 1.69, "learning_rate": 7.820764152830566e-06, "loss": 0.1319, "step": 5455 }, { "epoch": 1.69, "learning_rate": 7.820364072814564e-06, "loss": 0.2194, "step": 5456 }, { "epoch": 1.69, "learning_rate": 7.819963992798561e-06, "loss": 0.1913, "step": 5457 }, { "epoch": 1.69, "learning_rate": 7.819563912782558e-06, "loss": 0.2247, "step": 5458 }, { "epoch": 1.69, "learning_rate": 7.819163832766554e-06, "loss": 0.1792, "step": 5459 }, { "epoch": 1.69, "learning_rate": 7.818763752750551e-06, "loss": 0.1376, "step": 5460 }, { "epoch": 1.69, "learning_rate": 7.818363672734548e-06, "loss": 0.211, "step": 5461 }, { "epoch": 1.7, "learning_rate": 7.817963592718544e-06, "loss": 0.2428, "step": 5462 }, { "epoch": 1.7, "learning_rate": 7.817563512702541e-06, "loss": 0.133, "step": 5463 }, { "epoch": 1.7, "learning_rate": 7.817163432686538e-06, "loss": 0.1219, "step": 5464 }, { "epoch": 1.7, "learning_rate": 7.816763352670534e-06, "loss": 0.2588, "step": 5465 }, { "epoch": 1.7, "learning_rate": 7.816363272654531e-06, "loss": 0.0767, "step": 5466 }, { "epoch": 1.7, "learning_rate": 7.815963192638528e-06, "loss": 0.1332, "step": 5467 }, { "epoch": 1.7, "learning_rate": 7.815563112622525e-06, "loss": 0.2099, "step": 5468 }, { "epoch": 1.7, "learning_rate": 7.815163032606521e-06, "loss": 0.1669, "step": 5469 }, { "epoch": 1.7, "learning_rate": 7.81476295259052e-06, "loss": 0.1729, "step": 5470 }, { "epoch": 1.7, "learning_rate": 7.814362872574516e-06, "loss": 0.3439, "step": 5471 }, { "epoch": 1.7, "learning_rate": 7.813962792558513e-06, "loss": 0.3103, "step": 5472 }, { "epoch": 1.7, "learning_rate": 7.81356271254251e-06, "loss": 0.2013, "step": 5473 }, { "epoch": 1.7, "learning_rate": 7.813162632526506e-06, "loss": 0.1323, "step": 5474 }, { "epoch": 1.7, "learning_rate": 7.812762552510503e-06, "loss": 0.1282, "step": 5475 }, { "epoch": 1.7, "learning_rate": 7.8123624724945e-06, "loss": 0.1388, "step": 5476 }, { "epoch": 1.7, "learning_rate": 7.811962392478496e-06, "loss": 0.2316, "step": 5477 }, { "epoch": 1.7, "learning_rate": 7.811562312462493e-06, "loss": 0.0983, "step": 5478 }, { "epoch": 1.7, "learning_rate": 7.81116223244649e-06, "loss": 0.2334, "step": 5479 }, { "epoch": 1.7, "learning_rate": 7.810762152430486e-06, "loss": 0.1259, "step": 5480 }, { "epoch": 1.7, "learning_rate": 7.810362072414483e-06, "loss": 0.1402, "step": 5481 }, { "epoch": 1.7, "learning_rate": 7.80996199239848e-06, "loss": 0.2679, "step": 5482 }, { "epoch": 1.7, "learning_rate": 7.809561912382477e-06, "loss": 0.164, "step": 5483 }, { "epoch": 1.7, "learning_rate": 7.809161832366475e-06, "loss": 0.1754, "step": 5484 }, { "epoch": 1.7, "learning_rate": 7.808761752350472e-06, "loss": 0.0947, "step": 5485 }, { "epoch": 1.7, "learning_rate": 7.808361672334468e-06, "loss": 0.1743, "step": 5486 }, { "epoch": 1.7, "learning_rate": 7.807961592318465e-06, "loss": 0.104, "step": 5487 }, { "epoch": 1.7, "learning_rate": 7.80756151230246e-06, "loss": 0.1705, "step": 5488 }, { "epoch": 1.7, "learning_rate": 7.807161432286458e-06, "loss": 0.1274, "step": 5489 }, { "epoch": 1.7, "learning_rate": 7.806761352270455e-06, "loss": 0.1577, "step": 5490 }, { "epoch": 1.7, "learning_rate": 7.806361272254452e-06, "loss": 0.1374, "step": 5491 }, { "epoch": 1.7, "learning_rate": 7.805961192238448e-06, "loss": 0.159, "step": 5492 }, { "epoch": 1.7, "learning_rate": 7.805561112222445e-06, "loss": 0.143, "step": 5493 }, { "epoch": 1.71, "learning_rate": 7.805161032206442e-06, "loss": 0.1394, "step": 5494 }, { "epoch": 1.71, "learning_rate": 7.804760952190438e-06, "loss": 0.1856, "step": 5495 }, { "epoch": 1.71, "learning_rate": 7.804360872174435e-06, "loss": 0.2033, "step": 5496 }, { "epoch": 1.71, "learning_rate": 7.803960792158432e-06, "loss": 0.2011, "step": 5497 }, { "epoch": 1.71, "learning_rate": 7.80356071214243e-06, "loss": 0.4472, "step": 5498 }, { "epoch": 1.71, "learning_rate": 7.803160632126427e-06, "loss": 0.1661, "step": 5499 }, { "epoch": 1.71, "learning_rate": 7.802760552110424e-06, "loss": 0.1261, "step": 5500 }, { "epoch": 1.71, "learning_rate": 7.802360472094419e-06, "loss": 0.1071, "step": 5501 }, { "epoch": 1.71, "learning_rate": 7.801960392078415e-06, "loss": 0.0855, "step": 5502 }, { "epoch": 1.71, "learning_rate": 7.801560312062414e-06, "loss": 0.1664, "step": 5503 }, { "epoch": 1.71, "learning_rate": 7.80116023204641e-06, "loss": 0.1476, "step": 5504 }, { "epoch": 1.71, "learning_rate": 7.800760152030407e-06, "loss": 0.1141, "step": 5505 }, { "epoch": 1.71, "learning_rate": 7.800360072014404e-06, "loss": 0.2677, "step": 5506 }, { "epoch": 1.71, "learning_rate": 7.7999599919984e-06, "loss": 0.1272, "step": 5507 }, { "epoch": 1.71, "learning_rate": 7.799559911982397e-06, "loss": 0.0707, "step": 5508 }, { "epoch": 1.71, "learning_rate": 7.799159831966394e-06, "loss": 0.1083, "step": 5509 }, { "epoch": 1.71, "learning_rate": 7.79875975195039e-06, "loss": 0.2083, "step": 5510 }, { "epoch": 1.71, "learning_rate": 7.798359671934387e-06, "loss": 0.1692, "step": 5511 }, { "epoch": 1.71, "learning_rate": 7.797959591918385e-06, "loss": 0.1704, "step": 5512 }, { "epoch": 1.71, "learning_rate": 7.79755951190238e-06, "loss": 0.1357, "step": 5513 }, { "epoch": 1.71, "learning_rate": 7.797159431886377e-06, "loss": 0.1038, "step": 5514 }, { "epoch": 1.71, "learning_rate": 7.796759351870374e-06, "loss": 0.1853, "step": 5515 }, { "epoch": 1.71, "learning_rate": 7.79635927185437e-06, "loss": 0.2436, "step": 5516 }, { "epoch": 1.71, "learning_rate": 7.795959191838369e-06, "loss": 0.107, "step": 5517 }, { "epoch": 1.71, "learning_rate": 7.795559111822366e-06, "loss": 0.1728, "step": 5518 }, { "epoch": 1.71, "learning_rate": 7.795159031806362e-06, "loss": 0.1308, "step": 5519 }, { "epoch": 1.71, "learning_rate": 7.794758951790359e-06, "loss": 0.1506, "step": 5520 }, { "epoch": 1.71, "learning_rate": 7.794358871774356e-06, "loss": 0.2331, "step": 5521 }, { "epoch": 1.71, "learning_rate": 7.793958791758352e-06, "loss": 0.0856, "step": 5522 }, { "epoch": 1.71, "learning_rate": 7.793558711742349e-06, "loss": 0.244, "step": 5523 }, { "epoch": 1.71, "learning_rate": 7.793158631726346e-06, "loss": 0.2007, "step": 5524 }, { "epoch": 1.71, "learning_rate": 7.792758551710342e-06, "loss": 0.1103, "step": 5525 }, { "epoch": 1.72, "learning_rate": 7.792358471694339e-06, "loss": 0.1096, "step": 5526 }, { "epoch": 1.72, "learning_rate": 7.791958391678336e-06, "loss": 0.1369, "step": 5527 }, { "epoch": 1.72, "learning_rate": 7.791558311662332e-06, "loss": 0.1815, "step": 5528 }, { "epoch": 1.72, "learning_rate": 7.79115823164633e-06, "loss": 0.1293, "step": 5529 }, { "epoch": 1.72, "learning_rate": 7.790758151630326e-06, "loss": 0.1522, "step": 5530 }, { "epoch": 1.72, "learning_rate": 7.790358071614324e-06, "loss": 0.1354, "step": 5531 }, { "epoch": 1.72, "learning_rate": 7.789957991598321e-06, "loss": 0.1071, "step": 5532 }, { "epoch": 1.72, "learning_rate": 7.789557911582318e-06, "loss": 0.2335, "step": 5533 }, { "epoch": 1.72, "learning_rate": 7.789157831566314e-06, "loss": 0.2696, "step": 5534 }, { "epoch": 1.72, "learning_rate": 7.788757751550311e-06, "loss": 0.1151, "step": 5535 }, { "epoch": 1.72, "learning_rate": 7.788357671534308e-06, "loss": 0.142, "step": 5536 }, { "epoch": 1.72, "learning_rate": 7.787957591518304e-06, "loss": 0.1613, "step": 5537 }, { "epoch": 1.72, "learning_rate": 7.787557511502301e-06, "loss": 0.1903, "step": 5538 }, { "epoch": 1.72, "learning_rate": 7.787157431486298e-06, "loss": 0.0912, "step": 5539 }, { "epoch": 1.72, "learning_rate": 7.786757351470294e-06, "loss": 0.1237, "step": 5540 }, { "epoch": 1.72, "learning_rate": 7.786357271454291e-06, "loss": 0.0835, "step": 5541 }, { "epoch": 1.72, "learning_rate": 7.785957191438288e-06, "loss": 0.1453, "step": 5542 }, { "epoch": 1.72, "learning_rate": 7.785557111422284e-06, "loss": 0.1012, "step": 5543 }, { "epoch": 1.72, "learning_rate": 7.785157031406281e-06, "loss": 0.1601, "step": 5544 }, { "epoch": 1.72, "learning_rate": 7.78475695139028e-06, "loss": 0.0747, "step": 5545 }, { "epoch": 1.72, "learning_rate": 7.784356871374276e-06, "loss": 0.1657, "step": 5546 }, { "epoch": 1.72, "learning_rate": 7.783956791358273e-06, "loss": 0.1756, "step": 5547 }, { "epoch": 1.72, "learning_rate": 7.78355671134227e-06, "loss": 0.0694, "step": 5548 }, { "epoch": 1.72, "learning_rate": 7.783156631326266e-06, "loss": 0.1563, "step": 5549 }, { "epoch": 1.72, "learning_rate": 7.782756551310263e-06, "loss": 0.1704, "step": 5550 }, { "epoch": 1.72, "learning_rate": 7.78235647129426e-06, "loss": 0.143, "step": 5551 }, { "epoch": 1.72, "learning_rate": 7.781956391278256e-06, "loss": 0.2568, "step": 5552 }, { "epoch": 1.72, "learning_rate": 7.781556311262253e-06, "loss": 0.1508, "step": 5553 }, { "epoch": 1.72, "learning_rate": 7.78115623124625e-06, "loss": 0.0827, "step": 5554 }, { "epoch": 1.72, "learning_rate": 7.780756151230246e-06, "loss": 0.2165, "step": 5555 }, { "epoch": 1.72, "learning_rate": 7.780356071214243e-06, "loss": 0.1458, "step": 5556 }, { "epoch": 1.72, "learning_rate": 7.77995599119824e-06, "loss": 0.1839, "step": 5557 }, { "epoch": 1.73, "learning_rate": 7.779555911182236e-06, "loss": 0.0719, "step": 5558 }, { "epoch": 1.73, "learning_rate": 7.779155831166235e-06, "loss": 0.1261, "step": 5559 }, { "epoch": 1.73, "learning_rate": 7.778755751150232e-06, "loss": 0.2339, "step": 5560 }, { "epoch": 1.73, "learning_rate": 7.778355671134228e-06, "loss": 0.1745, "step": 5561 }, { "epoch": 1.73, "learning_rate": 7.777955591118225e-06, "loss": 0.278, "step": 5562 }, { "epoch": 1.73, "learning_rate": 7.777555511102222e-06, "loss": 0.1907, "step": 5563 }, { "epoch": 1.73, "learning_rate": 7.777155431086217e-06, "loss": 0.1215, "step": 5564 }, { "epoch": 1.73, "learning_rate": 7.776755351070215e-06, "loss": 0.1759, "step": 5565 }, { "epoch": 1.73, "learning_rate": 7.776355271054212e-06, "loss": 0.1867, "step": 5566 }, { "epoch": 1.73, "learning_rate": 7.775955191038208e-06, "loss": 0.3198, "step": 5567 }, { "epoch": 1.73, "learning_rate": 7.775555111022205e-06, "loss": 0.2255, "step": 5568 }, { "epoch": 1.73, "learning_rate": 7.775155031006202e-06, "loss": 0.1935, "step": 5569 }, { "epoch": 1.73, "learning_rate": 7.774754950990198e-06, "loss": 0.1421, "step": 5570 }, { "epoch": 1.73, "learning_rate": 7.774354870974195e-06, "loss": 0.0867, "step": 5571 }, { "epoch": 1.73, "learning_rate": 7.773954790958192e-06, "loss": 0.3858, "step": 5572 }, { "epoch": 1.73, "learning_rate": 7.77355471094219e-06, "loss": 0.1198, "step": 5573 }, { "epoch": 1.73, "learning_rate": 7.773154630926187e-06, "loss": 0.1051, "step": 5574 }, { "epoch": 1.73, "learning_rate": 7.772754550910184e-06, "loss": 0.1592, "step": 5575 }, { "epoch": 1.73, "learning_rate": 7.772354470894179e-06, "loss": 0.2174, "step": 5576 }, { "epoch": 1.73, "learning_rate": 7.771954390878175e-06, "loss": 0.1678, "step": 5577 }, { "epoch": 1.73, "learning_rate": 7.771554310862172e-06, "loss": 0.1867, "step": 5578 }, { "epoch": 1.73, "learning_rate": 7.77115423084617e-06, "loss": 0.1452, "step": 5579 }, { "epoch": 1.73, "learning_rate": 7.770754150830167e-06, "loss": 0.1546, "step": 5580 }, { "epoch": 1.73, "learning_rate": 7.770354070814164e-06, "loss": 0.1783, "step": 5581 }, { "epoch": 1.73, "learning_rate": 7.76995399079816e-06, "loss": 0.1257, "step": 5582 }, { "epoch": 1.73, "learning_rate": 7.769553910782157e-06, "loss": 0.1276, "step": 5583 }, { "epoch": 1.73, "learning_rate": 7.769153830766154e-06, "loss": 0.1339, "step": 5584 }, { "epoch": 1.73, "learning_rate": 7.76875375075015e-06, "loss": 0.1127, "step": 5585 }, { "epoch": 1.73, "learning_rate": 7.768353670734147e-06, "loss": 0.1832, "step": 5586 }, { "epoch": 1.73, "learning_rate": 7.767953590718145e-06, "loss": 0.137, "step": 5587 }, { "epoch": 1.73, "learning_rate": 7.767553510702142e-06, "loss": 0.1736, "step": 5588 }, { "epoch": 1.73, "learning_rate": 7.767153430686137e-06, "loss": 0.1159, "step": 5589 }, { "epoch": 1.74, "learning_rate": 7.766753350670134e-06, "loss": 0.2118, "step": 5590 }, { "epoch": 1.74, "learning_rate": 7.76635327065413e-06, "loss": 0.1878, "step": 5591 }, { "epoch": 1.74, "learning_rate": 7.765953190638129e-06, "loss": 0.1072, "step": 5592 }, { "epoch": 1.74, "learning_rate": 7.765553110622126e-06, "loss": 0.072, "step": 5593 }, { "epoch": 1.74, "learning_rate": 7.765153030606122e-06, "loss": 0.1704, "step": 5594 }, { "epoch": 1.74, "learning_rate": 7.764752950590119e-06, "loss": 0.2609, "step": 5595 }, { "epoch": 1.74, "learning_rate": 7.764352870574116e-06, "loss": 0.061, "step": 5596 }, { "epoch": 1.74, "learning_rate": 7.763952790558112e-06, "loss": 0.1267, "step": 5597 }, { "epoch": 1.74, "learning_rate": 7.763552710542109e-06, "loss": 0.0706, "step": 5598 }, { "epoch": 1.74, "learning_rate": 7.763152630526106e-06, "loss": 0.1556, "step": 5599 }, { "epoch": 1.74, "learning_rate": 7.762752550510102e-06, "loss": 0.2359, "step": 5600 }, { "epoch": 1.74, "learning_rate": 7.762352470494099e-06, "loss": 0.126, "step": 5601 }, { "epoch": 1.74, "learning_rate": 7.761952390478096e-06, "loss": 0.1172, "step": 5602 }, { "epoch": 1.74, "learning_rate": 7.761552310462092e-06, "loss": 0.1914, "step": 5603 }, { "epoch": 1.74, "learning_rate": 7.761152230446089e-06, "loss": 0.1043, "step": 5604 }, { "epoch": 1.74, "learning_rate": 7.760752150430086e-06, "loss": 0.1329, "step": 5605 }, { "epoch": 1.74, "learning_rate": 7.760352070414084e-06, "loss": 0.2344, "step": 5606 }, { "epoch": 1.74, "learning_rate": 7.759951990398081e-06, "loss": 0.0759, "step": 5607 }, { "epoch": 1.74, "learning_rate": 7.759551910382078e-06, "loss": 0.1328, "step": 5608 }, { "epoch": 1.74, "learning_rate": 7.759151830366074e-06, "loss": 0.143, "step": 5609 }, { "epoch": 1.74, "learning_rate": 7.758751750350071e-06, "loss": 0.1455, "step": 5610 }, { "epoch": 1.74, "learning_rate": 7.758351670334068e-06, "loss": 0.16, "step": 5611 }, { "epoch": 1.74, "learning_rate": 7.757951590318064e-06, "loss": 0.1951, "step": 5612 }, { "epoch": 1.74, "learning_rate": 7.757551510302061e-06, "loss": 0.1239, "step": 5613 }, { "epoch": 1.74, "learning_rate": 7.757151430286058e-06, "loss": 0.1756, "step": 5614 }, { "epoch": 1.74, "learning_rate": 7.756751350270054e-06, "loss": 0.1123, "step": 5615 }, { "epoch": 1.74, "learning_rate": 7.756351270254051e-06, "loss": 0.1219, "step": 5616 }, { "epoch": 1.74, "learning_rate": 7.755951190238048e-06, "loss": 0.1973, "step": 5617 }, { "epoch": 1.74, "learning_rate": 7.755551110222044e-06, "loss": 0.1566, "step": 5618 }, { "epoch": 1.74, "learning_rate": 7.755151030206041e-06, "loss": 0.2112, "step": 5619 }, { "epoch": 1.74, "learning_rate": 7.75475095019004e-06, "loss": 0.1282, "step": 5620 }, { "epoch": 1.74, "learning_rate": 7.754350870174036e-06, "loss": 0.2614, "step": 5621 }, { "epoch": 1.74, "learning_rate": 7.753950790158033e-06, "loss": 0.1305, "step": 5622 }, { "epoch": 1.75, "learning_rate": 7.75355071014203e-06, "loss": 0.1053, "step": 5623 }, { "epoch": 1.75, "learning_rate": 7.753150630126026e-06, "loss": 0.1715, "step": 5624 }, { "epoch": 1.75, "learning_rate": 7.752750550110023e-06, "loss": 0.1525, "step": 5625 }, { "epoch": 1.75, "learning_rate": 7.75235047009402e-06, "loss": 0.1061, "step": 5626 }, { "epoch": 1.75, "learning_rate": 7.751950390078016e-06, "loss": 0.1274, "step": 5627 }, { "epoch": 1.75, "learning_rate": 7.751550310062013e-06, "loss": 0.1316, "step": 5628 }, { "epoch": 1.75, "learning_rate": 7.75115023004601e-06, "loss": 0.1794, "step": 5629 }, { "epoch": 1.75, "learning_rate": 7.750750150030006e-06, "loss": 0.1233, "step": 5630 }, { "epoch": 1.75, "learning_rate": 7.750350070014003e-06, "loss": 0.0966, "step": 5631 }, { "epoch": 1.75, "learning_rate": 7.749949989998e-06, "loss": 0.1575, "step": 5632 }, { "epoch": 1.75, "learning_rate": 7.749549909981996e-06, "loss": 0.1471, "step": 5633 }, { "epoch": 1.75, "learning_rate": 7.749149829965995e-06, "loss": 0.1454, "step": 5634 }, { "epoch": 1.75, "learning_rate": 7.748749749949991e-06, "loss": 0.1326, "step": 5635 }, { "epoch": 1.75, "learning_rate": 7.748349669933988e-06, "loss": 0.1391, "step": 5636 }, { "epoch": 1.75, "learning_rate": 7.747949589917985e-06, "loss": 0.1396, "step": 5637 }, { "epoch": 1.75, "learning_rate": 7.747549509901982e-06, "loss": 0.1876, "step": 5638 }, { "epoch": 1.75, "learning_rate": 7.747149429885977e-06, "loss": 0.2574, "step": 5639 }, { "epoch": 1.75, "learning_rate": 7.746749349869975e-06, "loss": 0.1167, "step": 5640 }, { "epoch": 1.75, "learning_rate": 7.746349269853972e-06, "loss": 0.2221, "step": 5641 }, { "epoch": 1.75, "learning_rate": 7.745949189837968e-06, "loss": 0.196, "step": 5642 }, { "epoch": 1.75, "learning_rate": 7.745549109821965e-06, "loss": 0.1712, "step": 5643 }, { "epoch": 1.75, "learning_rate": 7.745149029805962e-06, "loss": 0.1044, "step": 5644 }, { "epoch": 1.75, "learning_rate": 7.744748949789958e-06, "loss": 0.1402, "step": 5645 }, { "epoch": 1.75, "learning_rate": 7.744348869773955e-06, "loss": 0.0616, "step": 5646 }, { "epoch": 1.75, "learning_rate": 7.743948789757952e-06, "loss": 0.2523, "step": 5647 }, { "epoch": 1.75, "learning_rate": 7.74354870974195e-06, "loss": 0.1498, "step": 5648 }, { "epoch": 1.75, "learning_rate": 7.743148629725947e-06, "loss": 0.3376, "step": 5649 }, { "epoch": 1.75, "learning_rate": 7.742748549709943e-06, "loss": 0.2615, "step": 5650 }, { "epoch": 1.75, "learning_rate": 7.74234846969394e-06, "loss": 0.141, "step": 5651 }, { "epoch": 1.75, "learning_rate": 7.741948389677935e-06, "loss": 0.1266, "step": 5652 }, { "epoch": 1.75, "learning_rate": 7.741548309661932e-06, "loss": 0.1551, "step": 5653 }, { "epoch": 1.75, "learning_rate": 7.74114822964593e-06, "loss": 0.1191, "step": 5654 }, { "epoch": 1.76, "learning_rate": 7.740748149629927e-06, "loss": 0.3383, "step": 5655 }, { "epoch": 1.76, "learning_rate": 7.740348069613924e-06, "loss": 0.1501, "step": 5656 }, { "epoch": 1.76, "learning_rate": 7.73994798959792e-06, "loss": 0.1439, "step": 5657 }, { "epoch": 1.76, "learning_rate": 7.739547909581917e-06, "loss": 0.1032, "step": 5658 }, { "epoch": 1.76, "learning_rate": 7.739147829565914e-06, "loss": 0.1232, "step": 5659 }, { "epoch": 1.76, "learning_rate": 7.73874774954991e-06, "loss": 0.1781, "step": 5660 }, { "epoch": 1.76, "learning_rate": 7.738347669533907e-06, "loss": 0.1825, "step": 5661 }, { "epoch": 1.76, "learning_rate": 7.737947589517905e-06, "loss": 0.1485, "step": 5662 }, { "epoch": 1.76, "learning_rate": 7.737547509501902e-06, "loss": 0.1886, "step": 5663 }, { "epoch": 1.76, "learning_rate": 7.737147429485899e-06, "loss": 0.1454, "step": 5664 }, { "epoch": 1.76, "learning_rate": 7.736747349469894e-06, "loss": 0.3068, "step": 5665 }, { "epoch": 1.76, "learning_rate": 7.73634726945389e-06, "loss": 0.325, "step": 5666 }, { "epoch": 1.76, "learning_rate": 7.735947189437887e-06, "loss": 0.2116, "step": 5667 }, { "epoch": 1.76, "learning_rate": 7.735547109421885e-06, "loss": 0.166, "step": 5668 }, { "epoch": 1.76, "learning_rate": 7.735147029405882e-06, "loss": 0.2138, "step": 5669 }, { "epoch": 1.76, "learning_rate": 7.734746949389879e-06, "loss": 0.1352, "step": 5670 }, { "epoch": 1.76, "learning_rate": 7.734346869373876e-06, "loss": 0.2764, "step": 5671 }, { "epoch": 1.76, "learning_rate": 7.733946789357872e-06, "loss": 0.204, "step": 5672 }, { "epoch": 1.76, "learning_rate": 7.733546709341869e-06, "loss": 0.0596, "step": 5673 }, { "epoch": 1.76, "learning_rate": 7.733146629325866e-06, "loss": 0.1334, "step": 5674 }, { "epoch": 1.76, "learning_rate": 7.732746549309862e-06, "loss": 0.1032, "step": 5675 }, { "epoch": 1.76, "learning_rate": 7.73234646929386e-06, "loss": 0.2651, "step": 5676 }, { "epoch": 1.76, "learning_rate": 7.731946389277856e-06, "loss": 0.1289, "step": 5677 }, { "epoch": 1.76, "learning_rate": 7.731546309261852e-06, "loss": 0.0797, "step": 5678 }, { "epoch": 1.76, "learning_rate": 7.731146229245849e-06, "loss": 0.1476, "step": 5679 }, { "epoch": 1.76, "learning_rate": 7.730746149229846e-06, "loss": 0.0758, "step": 5680 }, { "epoch": 1.76, "learning_rate": 7.730346069213842e-06, "loss": 0.2674, "step": 5681 }, { "epoch": 1.76, "learning_rate": 7.72994598919784e-06, "loss": 0.1396, "step": 5682 }, { "epoch": 1.76, "learning_rate": 7.729545909181837e-06, "loss": 0.1763, "step": 5683 }, { "epoch": 1.76, "learning_rate": 7.729145829165834e-06, "loss": 0.2161, "step": 5684 }, { "epoch": 1.76, "learning_rate": 7.728745749149831e-06, "loss": 0.1, "step": 5685 }, { "epoch": 1.76, "learning_rate": 7.728345669133828e-06, "loss": 0.1112, "step": 5686 }, { "epoch": 1.77, "learning_rate": 7.727945589117824e-06, "loss": 0.0993, "step": 5687 }, { "epoch": 1.77, "learning_rate": 7.727545509101821e-06, "loss": 0.1184, "step": 5688 }, { "epoch": 1.77, "learning_rate": 7.727145429085818e-06, "loss": 0.1569, "step": 5689 }, { "epoch": 1.77, "learning_rate": 7.726745349069814e-06, "loss": 0.2827, "step": 5690 }, { "epoch": 1.77, "learning_rate": 7.726345269053811e-06, "loss": 0.1064, "step": 5691 }, { "epoch": 1.77, "learning_rate": 7.725945189037808e-06, "loss": 0.1454, "step": 5692 }, { "epoch": 1.77, "learning_rate": 7.725545109021804e-06, "loss": 0.1801, "step": 5693 }, { "epoch": 1.77, "learning_rate": 7.725145029005801e-06, "loss": 0.1952, "step": 5694 }, { "epoch": 1.77, "learning_rate": 7.724744948989798e-06, "loss": 0.1941, "step": 5695 }, { "epoch": 1.77, "learning_rate": 7.724344868973796e-06, "loss": 0.1247, "step": 5696 }, { "epoch": 1.77, "learning_rate": 7.723944788957793e-06, "loss": 0.1536, "step": 5697 }, { "epoch": 1.77, "learning_rate": 7.72354470894179e-06, "loss": 0.223, "step": 5698 }, { "epoch": 1.77, "learning_rate": 7.723144628925786e-06, "loss": 0.1416, "step": 5699 }, { "epoch": 1.77, "learning_rate": 7.722744548909783e-06, "loss": 0.1585, "step": 5700 }, { "epoch": 1.77, "learning_rate": 7.72234446889378e-06, "loss": 0.0867, "step": 5701 }, { "epoch": 1.77, "learning_rate": 7.721944388877776e-06, "loss": 0.1738, "step": 5702 }, { "epoch": 1.77, "learning_rate": 7.721544308861773e-06, "loss": 0.1914, "step": 5703 }, { "epoch": 1.77, "learning_rate": 7.72114422884577e-06, "loss": 0.176, "step": 5704 }, { "epoch": 1.77, "learning_rate": 7.720744148829766e-06, "loss": 0.1333, "step": 5705 }, { "epoch": 1.77, "learning_rate": 7.720344068813763e-06, "loss": 0.1542, "step": 5706 }, { "epoch": 1.77, "learning_rate": 7.71994398879776e-06, "loss": 0.1166, "step": 5707 }, { "epoch": 1.77, "learning_rate": 7.719543908781756e-06, "loss": 0.3547, "step": 5708 }, { "epoch": 1.77, "learning_rate": 7.719143828765753e-06, "loss": 0.1929, "step": 5709 }, { "epoch": 1.77, "learning_rate": 7.718743748749751e-06, "loss": 0.1958, "step": 5710 }, { "epoch": 1.77, "learning_rate": 7.718343668733748e-06, "loss": 0.1041, "step": 5711 }, { "epoch": 1.77, "learning_rate": 7.717943588717745e-06, "loss": 0.2035, "step": 5712 }, { "epoch": 1.77, "learning_rate": 7.717543508701741e-06, "loss": 0.063, "step": 5713 }, { "epoch": 1.77, "learning_rate": 7.717143428685738e-06, "loss": 0.1106, "step": 5714 }, { "epoch": 1.77, "learning_rate": 7.716743348669735e-06, "loss": 0.1962, "step": 5715 }, { "epoch": 1.77, "learning_rate": 7.716343268653732e-06, "loss": 0.2346, "step": 5716 }, { "epoch": 1.77, "learning_rate": 7.715943188637728e-06, "loss": 0.1494, "step": 5717 }, { "epoch": 1.77, "learning_rate": 7.715543108621725e-06, "loss": 0.1302, "step": 5718 }, { "epoch": 1.78, "learning_rate": 7.715143028605722e-06, "loss": 0.1264, "step": 5719 }, { "epoch": 1.78, "learning_rate": 7.714742948589718e-06, "loss": 0.2133, "step": 5720 }, { "epoch": 1.78, "learning_rate": 7.714342868573715e-06, "loss": 0.1711, "step": 5721 }, { "epoch": 1.78, "learning_rate": 7.713942788557712e-06, "loss": 0.2538, "step": 5722 }, { "epoch": 1.78, "learning_rate": 7.713542708541708e-06, "loss": 0.172, "step": 5723 }, { "epoch": 1.78, "learning_rate": 7.713142628525707e-06, "loss": 0.1446, "step": 5724 }, { "epoch": 1.78, "learning_rate": 7.712742548509703e-06, "loss": 0.1956, "step": 5725 }, { "epoch": 1.78, "learning_rate": 7.7123424684937e-06, "loss": 0.1282, "step": 5726 }, { "epoch": 1.78, "learning_rate": 7.711942388477697e-06, "loss": 0.1524, "step": 5727 }, { "epoch": 1.78, "learning_rate": 7.711542308461692e-06, "loss": 0.1621, "step": 5728 }, { "epoch": 1.78, "learning_rate": 7.71114222844569e-06, "loss": 0.1876, "step": 5729 }, { "epoch": 1.78, "learning_rate": 7.710742148429687e-06, "loss": 0.2425, "step": 5730 }, { "epoch": 1.78, "learning_rate": 7.710342068413684e-06, "loss": 0.0884, "step": 5731 }, { "epoch": 1.78, "learning_rate": 7.70994198839768e-06, "loss": 0.1754, "step": 5732 }, { "epoch": 1.78, "learning_rate": 7.709541908381677e-06, "loss": 0.164, "step": 5733 }, { "epoch": 1.78, "learning_rate": 7.709141828365674e-06, "loss": 0.1783, "step": 5734 }, { "epoch": 1.78, "learning_rate": 7.70874174834967e-06, "loss": 0.1502, "step": 5735 }, { "epoch": 1.78, "learning_rate": 7.708341668333667e-06, "loss": 0.117, "step": 5736 }, { "epoch": 1.78, "learning_rate": 7.707941588317665e-06, "loss": 0.1382, "step": 5737 }, { "epoch": 1.78, "learning_rate": 7.707541508301662e-06, "loss": 0.1137, "step": 5738 }, { "epoch": 1.78, "learning_rate": 7.707141428285659e-06, "loss": 0.1491, "step": 5739 }, { "epoch": 1.78, "learning_rate": 7.706741348269654e-06, "loss": 0.1991, "step": 5740 }, { "epoch": 1.78, "learning_rate": 7.70634126825365e-06, "loss": 0.071, "step": 5741 }, { "epoch": 1.78, "learning_rate": 7.705941188237647e-06, "loss": 0.2126, "step": 5742 }, { "epoch": 1.78, "learning_rate": 7.705541108221645e-06, "loss": 0.1494, "step": 5743 }, { "epoch": 1.78, "learning_rate": 7.705141028205642e-06, "loss": 0.1922, "step": 5744 }, { "epoch": 1.78, "learning_rate": 7.704740948189639e-06, "loss": 0.1738, "step": 5745 }, { "epoch": 1.78, "learning_rate": 7.704340868173635e-06, "loss": 0.156, "step": 5746 }, { "epoch": 1.78, "learning_rate": 7.703940788157632e-06, "loss": 0.1598, "step": 5747 }, { "epoch": 1.78, "learning_rate": 7.703540708141629e-06, "loss": 0.1991, "step": 5748 }, { "epoch": 1.78, "learning_rate": 7.703140628125626e-06, "loss": 0.2496, "step": 5749 }, { "epoch": 1.78, "learning_rate": 7.702740548109622e-06, "loss": 0.1833, "step": 5750 }, { "epoch": 1.78, "learning_rate": 7.70234046809362e-06, "loss": 0.1308, "step": 5751 }, { "epoch": 1.79, "learning_rate": 7.701940388077617e-06, "loss": 0.0869, "step": 5752 }, { "epoch": 1.79, "learning_rate": 7.701540308061612e-06, "loss": 0.2089, "step": 5753 }, { "epoch": 1.79, "learning_rate": 7.701140228045609e-06, "loss": 0.1927, "step": 5754 }, { "epoch": 1.79, "learning_rate": 7.700740148029606e-06, "loss": 0.1611, "step": 5755 }, { "epoch": 1.79, "learning_rate": 7.700340068013602e-06, "loss": 0.1689, "step": 5756 }, { "epoch": 1.79, "learning_rate": 7.6999399879976e-06, "loss": 0.1368, "step": 5757 }, { "epoch": 1.79, "learning_rate": 7.699539907981597e-06, "loss": 0.284, "step": 5758 }, { "epoch": 1.79, "learning_rate": 7.699139827965594e-06, "loss": 0.1849, "step": 5759 }, { "epoch": 1.79, "learning_rate": 7.69873974794959e-06, "loss": 0.179, "step": 5760 }, { "epoch": 1.79, "learning_rate": 7.698339667933587e-06, "loss": 0.2187, "step": 5761 }, { "epoch": 1.79, "learning_rate": 7.697939587917584e-06, "loss": 0.1065, "step": 5762 }, { "epoch": 1.79, "learning_rate": 7.697539507901581e-06, "loss": 0.1664, "step": 5763 }, { "epoch": 1.79, "learning_rate": 7.697139427885578e-06, "loss": 0.1481, "step": 5764 }, { "epoch": 1.79, "learning_rate": 7.696739347869574e-06, "loss": 0.0799, "step": 5765 }, { "epoch": 1.79, "learning_rate": 7.696339267853571e-06, "loss": 0.1434, "step": 5766 }, { "epoch": 1.79, "learning_rate": 7.695939187837568e-06, "loss": 0.1074, "step": 5767 }, { "epoch": 1.79, "learning_rate": 7.695539107821564e-06, "loss": 0.2006, "step": 5768 }, { "epoch": 1.79, "learning_rate": 7.695139027805561e-06, "loss": 0.0865, "step": 5769 }, { "epoch": 1.79, "learning_rate": 7.694738947789558e-06, "loss": 0.118, "step": 5770 }, { "epoch": 1.79, "learning_rate": 7.694338867773556e-06, "loss": 0.1692, "step": 5771 }, { "epoch": 1.79, "learning_rate": 7.693938787757553e-06, "loss": 0.1149, "step": 5772 }, { "epoch": 1.79, "learning_rate": 7.69353870774155e-06, "loss": 0.1045, "step": 5773 }, { "epoch": 1.79, "learning_rate": 7.693138627725546e-06, "loss": 0.0978, "step": 5774 }, { "epoch": 1.79, "learning_rate": 7.692738547709543e-06, "loss": 0.1946, "step": 5775 }, { "epoch": 1.79, "learning_rate": 7.69233846769354e-06, "loss": 0.1479, "step": 5776 }, { "epoch": 1.79, "learning_rate": 7.691938387677536e-06, "loss": 0.1215, "step": 5777 }, { "epoch": 1.79, "learning_rate": 7.691538307661533e-06, "loss": 0.0591, "step": 5778 }, { "epoch": 1.79, "learning_rate": 7.69113822764553e-06, "loss": 0.1547, "step": 5779 }, { "epoch": 1.79, "learning_rate": 7.690738147629526e-06, "loss": 0.0921, "step": 5780 }, { "epoch": 1.79, "learning_rate": 7.690338067613523e-06, "loss": 0.288, "step": 5781 }, { "epoch": 1.79, "learning_rate": 7.68993798759752e-06, "loss": 0.2114, "step": 5782 }, { "epoch": 1.79, "learning_rate": 7.689537907581516e-06, "loss": 0.227, "step": 5783 }, { "epoch": 1.8, "learning_rate": 7.689137827565513e-06, "loss": 0.2883, "step": 5784 }, { "epoch": 1.8, "learning_rate": 7.688737747549511e-06, "loss": 0.1068, "step": 5785 }, { "epoch": 1.8, "learning_rate": 7.688337667533508e-06, "loss": 0.1478, "step": 5786 }, { "epoch": 1.8, "learning_rate": 7.687937587517505e-06, "loss": 0.1266, "step": 5787 }, { "epoch": 1.8, "learning_rate": 7.687537507501501e-06, "loss": 0.1451, "step": 5788 }, { "epoch": 1.8, "learning_rate": 7.687137427485498e-06, "loss": 0.1102, "step": 5789 }, { "epoch": 1.8, "learning_rate": 7.686737347469495e-06, "loss": 0.15, "step": 5790 }, { "epoch": 1.8, "learning_rate": 7.686337267453491e-06, "loss": 0.1314, "step": 5791 }, { "epoch": 1.8, "learning_rate": 7.685937187437488e-06, "loss": 0.173, "step": 5792 }, { "epoch": 1.8, "learning_rate": 7.685537107421485e-06, "loss": 0.0799, "step": 5793 }, { "epoch": 1.8, "learning_rate": 7.685137027405482e-06, "loss": 0.1237, "step": 5794 }, { "epoch": 1.8, "learning_rate": 7.684736947389478e-06, "loss": 0.1703, "step": 5795 }, { "epoch": 1.8, "learning_rate": 7.684336867373475e-06, "loss": 0.1607, "step": 5796 }, { "epoch": 1.8, "learning_rate": 7.683936787357472e-06, "loss": 0.1434, "step": 5797 }, { "epoch": 1.8, "learning_rate": 7.683536707341468e-06, "loss": 0.2835, "step": 5798 }, { "epoch": 1.8, "learning_rate": 7.683136627325467e-06, "loss": 0.1668, "step": 5799 }, { "epoch": 1.8, "learning_rate": 7.682736547309463e-06, "loss": 0.1083, "step": 5800 }, { "epoch": 1.8, "learning_rate": 7.68233646729346e-06, "loss": 0.1497, "step": 5801 }, { "epoch": 1.8, "learning_rate": 7.681936387277457e-06, "loss": 0.2332, "step": 5802 }, { "epoch": 1.8, "learning_rate": 7.681536307261452e-06, "loss": 0.1748, "step": 5803 }, { "epoch": 1.8, "learning_rate": 7.68113622724545e-06, "loss": 0.2778, "step": 5804 }, { "epoch": 1.8, "learning_rate": 7.680736147229447e-06, "loss": 0.1135, "step": 5805 }, { "epoch": 1.8, "learning_rate": 7.680336067213443e-06, "loss": 0.1672, "step": 5806 }, { "epoch": 1.8, "learning_rate": 7.67993598719744e-06, "loss": 0.1792, "step": 5807 }, { "epoch": 1.8, "learning_rate": 7.679535907181437e-06, "loss": 0.0444, "step": 5808 }, { "epoch": 1.8, "learning_rate": 7.679135827165434e-06, "loss": 0.2265, "step": 5809 }, { "epoch": 1.8, "learning_rate": 7.67873574714943e-06, "loss": 0.0784, "step": 5810 }, { "epoch": 1.8, "learning_rate": 7.678335667133427e-06, "loss": 0.2362, "step": 5811 }, { "epoch": 1.8, "learning_rate": 7.677935587117424e-06, "loss": 0.1378, "step": 5812 }, { "epoch": 1.8, "learning_rate": 7.677535507101422e-06, "loss": 0.1178, "step": 5813 }, { "epoch": 1.8, "learning_rate": 7.677135427085419e-06, "loss": 0.0781, "step": 5814 }, { "epoch": 1.8, "learning_rate": 7.676735347069415e-06, "loss": 0.1047, "step": 5815 }, { "epoch": 1.81, "learning_rate": 7.67633526705341e-06, "loss": 0.1272, "step": 5816 }, { "epoch": 1.81, "learning_rate": 7.675935187037407e-06, "loss": 0.1878, "step": 5817 }, { "epoch": 1.81, "learning_rate": 7.675535107021405e-06, "loss": 0.1719, "step": 5818 }, { "epoch": 1.81, "learning_rate": 7.675135027005402e-06, "loss": 0.1071, "step": 5819 }, { "epoch": 1.81, "learning_rate": 7.674734946989399e-06, "loss": 0.1974, "step": 5820 }, { "epoch": 1.81, "learning_rate": 7.674334866973395e-06, "loss": 0.2102, "step": 5821 }, { "epoch": 1.81, "learning_rate": 7.673934786957392e-06, "loss": 0.1073, "step": 5822 }, { "epoch": 1.81, "learning_rate": 7.673534706941389e-06, "loss": 0.2229, "step": 5823 }, { "epoch": 1.81, "learning_rate": 7.673134626925385e-06, "loss": 0.1186, "step": 5824 }, { "epoch": 1.81, "learning_rate": 7.672734546909382e-06, "loss": 0.1895, "step": 5825 }, { "epoch": 1.81, "learning_rate": 7.672334466893379e-06, "loss": 0.1095, "step": 5826 }, { "epoch": 1.81, "learning_rate": 7.671934386877377e-06, "loss": 0.1969, "step": 5827 }, { "epoch": 1.81, "learning_rate": 7.671534306861372e-06, "loss": 0.1675, "step": 5828 }, { "epoch": 1.81, "learning_rate": 7.671134226845369e-06, "loss": 0.1805, "step": 5829 }, { "epoch": 1.81, "learning_rate": 7.670734146829366e-06, "loss": 0.0897, "step": 5830 }, { "epoch": 1.81, "learning_rate": 7.670334066813362e-06, "loss": 0.1135, "step": 5831 }, { "epoch": 1.81, "learning_rate": 7.66993398679736e-06, "loss": 0.2169, "step": 5832 }, { "epoch": 1.81, "learning_rate": 7.669533906781357e-06, "loss": 0.0903, "step": 5833 }, { "epoch": 1.81, "learning_rate": 7.669133826765354e-06, "loss": 0.142, "step": 5834 }, { "epoch": 1.81, "learning_rate": 7.66873374674935e-06, "loss": 0.1662, "step": 5835 }, { "epoch": 1.81, "learning_rate": 7.668333666733347e-06, "loss": 0.1177, "step": 5836 }, { "epoch": 1.81, "learning_rate": 7.667933586717344e-06, "loss": 0.1526, "step": 5837 }, { "epoch": 1.81, "learning_rate": 7.66753350670134e-06, "loss": 0.1086, "step": 5838 }, { "epoch": 1.81, "learning_rate": 7.667133426685337e-06, "loss": 0.1309, "step": 5839 }, { "epoch": 1.81, "learning_rate": 7.666733346669334e-06, "loss": 0.1122, "step": 5840 }, { "epoch": 1.81, "learning_rate": 7.666333266653331e-06, "loss": 0.1374, "step": 5841 }, { "epoch": 1.81, "learning_rate": 7.665933186637328e-06, "loss": 0.1718, "step": 5842 }, { "epoch": 1.81, "learning_rate": 7.665533106621324e-06, "loss": 0.2253, "step": 5843 }, { "epoch": 1.81, "learning_rate": 7.665133026605321e-06, "loss": 0.1272, "step": 5844 }, { "epoch": 1.81, "learning_rate": 7.664732946589318e-06, "loss": 0.1773, "step": 5845 }, { "epoch": 1.81, "learning_rate": 7.664332866573316e-06, "loss": 0.3166, "step": 5846 }, { "epoch": 1.81, "learning_rate": 7.663932786557313e-06, "loss": 0.0775, "step": 5847 }, { "epoch": 1.82, "learning_rate": 7.66353270654131e-06, "loss": 0.1138, "step": 5848 }, { "epoch": 1.82, "learning_rate": 7.663132626525306e-06, "loss": 0.1498, "step": 5849 }, { "epoch": 1.82, "learning_rate": 7.662732546509303e-06, "loss": 0.2077, "step": 5850 }, { "epoch": 1.82, "learning_rate": 7.6623324664933e-06, "loss": 0.0946, "step": 5851 }, { "epoch": 1.82, "learning_rate": 7.661932386477296e-06, "loss": 0.1205, "step": 5852 }, { "epoch": 1.82, "learning_rate": 7.661532306461293e-06, "loss": 0.1408, "step": 5853 }, { "epoch": 1.82, "learning_rate": 7.66113222644529e-06, "loss": 0.0943, "step": 5854 }, { "epoch": 1.82, "learning_rate": 7.660732146429286e-06, "loss": 0.1362, "step": 5855 }, { "epoch": 1.82, "learning_rate": 7.660332066413283e-06, "loss": 0.1244, "step": 5856 }, { "epoch": 1.82, "learning_rate": 7.65993198639728e-06, "loss": 0.1915, "step": 5857 }, { "epoch": 1.82, "learning_rate": 7.659531906381276e-06, "loss": 0.0869, "step": 5858 }, { "epoch": 1.82, "learning_rate": 7.659131826365273e-06, "loss": 0.1458, "step": 5859 }, { "epoch": 1.82, "learning_rate": 7.658731746349271e-06, "loss": 0.1527, "step": 5860 }, { "epoch": 1.82, "learning_rate": 7.658331666333268e-06, "loss": 0.0904, "step": 5861 }, { "epoch": 1.82, "learning_rate": 7.657931586317265e-06, "loss": 0.1898, "step": 5862 }, { "epoch": 1.82, "learning_rate": 7.657531506301261e-06, "loss": 0.1531, "step": 5863 }, { "epoch": 1.82, "learning_rate": 7.657131426285258e-06, "loss": 0.1579, "step": 5864 }, { "epoch": 1.82, "learning_rate": 7.656731346269255e-06, "loss": 0.2194, "step": 5865 }, { "epoch": 1.82, "learning_rate": 7.656331266253251e-06, "loss": 0.0957, "step": 5866 }, { "epoch": 1.82, "learning_rate": 7.655931186237248e-06, "loss": 0.0675, "step": 5867 }, { "epoch": 1.82, "learning_rate": 7.655531106221245e-06, "loss": 0.1998, "step": 5868 }, { "epoch": 1.82, "learning_rate": 7.655131026205241e-06, "loss": 0.1734, "step": 5869 }, { "epoch": 1.82, "learning_rate": 7.654730946189238e-06, "loss": 0.1309, "step": 5870 }, { "epoch": 1.82, "learning_rate": 7.654330866173235e-06, "loss": 0.1059, "step": 5871 }, { "epoch": 1.82, "learning_rate": 7.653930786157232e-06, "loss": 0.0936, "step": 5872 }, { "epoch": 1.82, "learning_rate": 7.653530706141228e-06, "loss": 0.0642, "step": 5873 }, { "epoch": 1.82, "learning_rate": 7.653130626125227e-06, "loss": 0.1692, "step": 5874 }, { "epoch": 1.82, "learning_rate": 7.652730546109223e-06, "loss": 0.1528, "step": 5875 }, { "epoch": 1.82, "learning_rate": 7.65233046609322e-06, "loss": 0.1248, "step": 5876 }, { "epoch": 1.82, "learning_rate": 7.651930386077217e-06, "loss": 0.1741, "step": 5877 }, { "epoch": 1.82, "learning_rate": 7.651530306061213e-06, "loss": 0.1379, "step": 5878 }, { "epoch": 1.82, "learning_rate": 7.651130226045208e-06, "loss": 0.2205, "step": 5879 }, { "epoch": 1.83, "learning_rate": 7.650730146029207e-06, "loss": 0.1641, "step": 5880 }, { "epoch": 1.83, "learning_rate": 7.650330066013203e-06, "loss": 0.1179, "step": 5881 }, { "epoch": 1.83, "learning_rate": 7.6499299859972e-06, "loss": 0.2815, "step": 5882 }, { "epoch": 1.83, "learning_rate": 7.649529905981197e-06, "loss": 0.113, "step": 5883 }, { "epoch": 1.83, "learning_rate": 7.649129825965193e-06, "loss": 0.1524, "step": 5884 }, { "epoch": 1.83, "learning_rate": 7.64872974594919e-06, "loss": 0.072, "step": 5885 }, { "epoch": 1.83, "learning_rate": 7.648329665933187e-06, "loss": 0.0662, "step": 5886 }, { "epoch": 1.83, "learning_rate": 7.647929585917184e-06, "loss": 0.2833, "step": 5887 }, { "epoch": 1.83, "learning_rate": 7.647529505901182e-06, "loss": 0.1747, "step": 5888 }, { "epoch": 1.83, "learning_rate": 7.647129425885179e-06, "loss": 0.135, "step": 5889 }, { "epoch": 1.83, "learning_rate": 7.646729345869175e-06, "loss": 0.187, "step": 5890 }, { "epoch": 1.83, "learning_rate": 7.646329265853172e-06, "loss": 0.1282, "step": 5891 }, { "epoch": 1.83, "learning_rate": 7.645929185837167e-06, "loss": 0.1129, "step": 5892 }, { "epoch": 1.83, "learning_rate": 7.645529105821165e-06, "loss": 0.1287, "step": 5893 }, { "epoch": 1.83, "learning_rate": 7.645129025805162e-06, "loss": 0.0767, "step": 5894 }, { "epoch": 1.83, "learning_rate": 7.644728945789159e-06, "loss": 0.1417, "step": 5895 }, { "epoch": 1.83, "learning_rate": 7.644328865773155e-06, "loss": 0.1547, "step": 5896 }, { "epoch": 1.83, "learning_rate": 7.643928785757152e-06, "loss": 0.1298, "step": 5897 }, { "epoch": 1.83, "learning_rate": 7.643528705741149e-06, "loss": 0.26, "step": 5898 }, { "epoch": 1.83, "learning_rate": 7.643128625725145e-06, "loss": 0.224, "step": 5899 }, { "epoch": 1.83, "learning_rate": 7.642728545709142e-06, "loss": 0.1178, "step": 5900 }, { "epoch": 1.83, "learning_rate": 7.642328465693139e-06, "loss": 0.2105, "step": 5901 }, { "epoch": 1.83, "learning_rate": 7.641928385677137e-06, "loss": 0.1212, "step": 5902 }, { "epoch": 1.83, "learning_rate": 7.641528305661134e-06, "loss": 0.2231, "step": 5903 }, { "epoch": 1.83, "learning_rate": 7.641128225645129e-06, "loss": 0.1301, "step": 5904 }, { "epoch": 1.83, "learning_rate": 7.640728145629126e-06, "loss": 0.1101, "step": 5905 }, { "epoch": 1.83, "learning_rate": 7.640328065613122e-06, "loss": 0.1473, "step": 5906 }, { "epoch": 1.83, "learning_rate": 7.63992798559712e-06, "loss": 0.1275, "step": 5907 }, { "epoch": 1.83, "learning_rate": 7.639527905581117e-06, "loss": 0.0992, "step": 5908 }, { "epoch": 1.83, "learning_rate": 7.639127825565114e-06, "loss": 0.216, "step": 5909 }, { "epoch": 1.83, "learning_rate": 7.63872774554911e-06, "loss": 0.1063, "step": 5910 }, { "epoch": 1.83, "learning_rate": 7.638327665533107e-06, "loss": 0.1552, "step": 5911 }, { "epoch": 1.83, "learning_rate": 7.637927585517104e-06, "loss": 0.1156, "step": 5912 }, { "epoch": 1.84, "learning_rate": 7.6375275055011e-06, "loss": 0.304, "step": 5913 }, { "epoch": 1.84, "learning_rate": 7.637127425485097e-06, "loss": 0.15, "step": 5914 }, { "epoch": 1.84, "learning_rate": 7.636727345469094e-06, "loss": 0.1278, "step": 5915 }, { "epoch": 1.84, "learning_rate": 7.636327265453092e-06, "loss": 0.1548, "step": 5916 }, { "epoch": 1.84, "learning_rate": 7.635927185437087e-06, "loss": 0.1513, "step": 5917 }, { "epoch": 1.84, "learning_rate": 7.635527105421084e-06, "loss": 0.151, "step": 5918 }, { "epoch": 1.84, "learning_rate": 7.635127025405081e-06, "loss": 0.0938, "step": 5919 }, { "epoch": 1.84, "learning_rate": 7.634726945389078e-06, "loss": 0.2661, "step": 5920 }, { "epoch": 1.84, "learning_rate": 7.634326865373076e-06, "loss": 0.0964, "step": 5921 }, { "epoch": 1.84, "learning_rate": 7.633926785357073e-06, "loss": 0.1339, "step": 5922 }, { "epoch": 1.84, "learning_rate": 7.63352670534107e-06, "loss": 0.1589, "step": 5923 }, { "epoch": 1.84, "learning_rate": 7.633126625325066e-06, "loss": 0.1792, "step": 5924 }, { "epoch": 1.84, "learning_rate": 7.632726545309063e-06, "loss": 0.1957, "step": 5925 }, { "epoch": 1.84, "learning_rate": 7.63232646529306e-06, "loss": 0.1028, "step": 5926 }, { "epoch": 1.84, "learning_rate": 7.631926385277056e-06, "loss": 0.1367, "step": 5927 }, { "epoch": 1.84, "learning_rate": 7.631526305261053e-06, "loss": 0.1754, "step": 5928 }, { "epoch": 1.84, "learning_rate": 7.63112622524505e-06, "loss": 0.1439, "step": 5929 }, { "epoch": 1.84, "learning_rate": 7.630726145229046e-06, "loss": 0.124, "step": 5930 }, { "epoch": 1.84, "learning_rate": 7.630326065213043e-06, "loss": 0.1669, "step": 5931 }, { "epoch": 1.84, "learning_rate": 7.62992598519704e-06, "loss": 0.1618, "step": 5932 }, { "epoch": 1.84, "learning_rate": 7.629525905181036e-06, "loss": 0.2757, "step": 5933 }, { "epoch": 1.84, "learning_rate": 7.629125825165034e-06, "loss": 0.1264, "step": 5934 }, { "epoch": 1.84, "learning_rate": 7.62872574514903e-06, "loss": 0.1264, "step": 5935 }, { "epoch": 1.84, "learning_rate": 7.628325665133027e-06, "loss": 0.2744, "step": 5936 }, { "epoch": 1.84, "learning_rate": 7.627925585117025e-06, "loss": 0.117, "step": 5937 }, { "epoch": 1.84, "learning_rate": 7.627525505101021e-06, "loss": 0.1062, "step": 5938 }, { "epoch": 1.84, "learning_rate": 7.627125425085018e-06, "loss": 0.1178, "step": 5939 }, { "epoch": 1.84, "learning_rate": 7.626725345069015e-06, "loss": 0.0594, "step": 5940 }, { "epoch": 1.84, "learning_rate": 7.626325265053011e-06, "loss": 0.1709, "step": 5941 }, { "epoch": 1.84, "learning_rate": 7.625925185037007e-06, "loss": 0.18, "step": 5942 }, { "epoch": 1.84, "learning_rate": 7.625525105021005e-06, "loss": 0.2155, "step": 5943 }, { "epoch": 1.84, "learning_rate": 7.625125025005001e-06, "loss": 0.1109, "step": 5944 }, { "epoch": 1.85, "learning_rate": 7.624724944988998e-06, "loss": 0.11, "step": 5945 }, { "epoch": 1.85, "learning_rate": 7.624324864972995e-06, "loss": 0.1072, "step": 5946 }, { "epoch": 1.85, "learning_rate": 7.623924784956992e-06, "loss": 0.1427, "step": 5947 }, { "epoch": 1.85, "learning_rate": 7.623524704940989e-06, "loss": 0.1024, "step": 5948 }, { "epoch": 1.85, "learning_rate": 7.623124624924986e-06, "loss": 0.156, "step": 5949 }, { "epoch": 1.85, "learning_rate": 7.622724544908982e-06, "loss": 0.1333, "step": 5950 }, { "epoch": 1.85, "learning_rate": 7.62232446489298e-06, "loss": 0.2082, "step": 5951 }, { "epoch": 1.85, "learning_rate": 7.621924384876977e-06, "loss": 0.161, "step": 5952 }, { "epoch": 1.85, "learning_rate": 7.621524304860973e-06, "loss": 0.1814, "step": 5953 }, { "epoch": 1.85, "learning_rate": 7.62112422484497e-06, "loss": 0.1171, "step": 5954 }, { "epoch": 1.85, "learning_rate": 7.620724144828966e-06, "loss": 0.2586, "step": 5955 }, { "epoch": 1.85, "learning_rate": 7.6203240648129625e-06, "loss": 0.0796, "step": 5956 }, { "epoch": 1.85, "learning_rate": 7.61992398479696e-06, "loss": 0.168, "step": 5957 }, { "epoch": 1.85, "learning_rate": 7.619523904780957e-06, "loss": 0.1647, "step": 5958 }, { "epoch": 1.85, "learning_rate": 7.619123824764953e-06, "loss": 0.1246, "step": 5959 }, { "epoch": 1.85, "learning_rate": 7.61872374474895e-06, "loss": 0.3151, "step": 5960 }, { "epoch": 1.85, "learning_rate": 7.618323664732948e-06, "loss": 0.1304, "step": 5961 }, { "epoch": 1.85, "learning_rate": 7.617923584716944e-06, "loss": 0.1221, "step": 5962 }, { "epoch": 1.85, "learning_rate": 7.617523504700941e-06, "loss": 0.1798, "step": 5963 }, { "epoch": 1.85, "learning_rate": 7.617123424684938e-06, "loss": 0.1892, "step": 5964 }, { "epoch": 1.85, "learning_rate": 7.616723344668935e-06, "loss": 0.1232, "step": 5965 }, { "epoch": 1.85, "learning_rate": 7.616323264652932e-06, "loss": 0.2158, "step": 5966 }, { "epoch": 1.85, "learning_rate": 7.615923184636928e-06, "loss": 0.1223, "step": 5967 }, { "epoch": 1.85, "learning_rate": 7.615523104620924e-06, "loss": 0.1838, "step": 5968 }, { "epoch": 1.85, "learning_rate": 7.615123024604921e-06, "loss": 0.1018, "step": 5969 }, { "epoch": 1.85, "learning_rate": 7.614722944588918e-06, "loss": 0.1809, "step": 5970 }, { "epoch": 1.85, "learning_rate": 7.614322864572915e-06, "loss": 0.1209, "step": 5971 }, { "epoch": 1.85, "learning_rate": 7.613922784556912e-06, "loss": 0.1163, "step": 5972 }, { "epoch": 1.85, "learning_rate": 7.613522704540909e-06, "loss": 0.1396, "step": 5973 }, { "epoch": 1.85, "learning_rate": 7.613122624524905e-06, "loss": 0.2142, "step": 5974 }, { "epoch": 1.85, "learning_rate": 7.612722544508903e-06, "loss": 0.1797, "step": 5975 }, { "epoch": 1.85, "learning_rate": 7.6123224644929e-06, "loss": 0.2099, "step": 5976 }, { "epoch": 1.86, "learning_rate": 7.611922384476896e-06, "loss": 0.1274, "step": 5977 }, { "epoch": 1.86, "learning_rate": 7.611522304460893e-06, "loss": 0.212, "step": 5978 }, { "epoch": 1.86, "learning_rate": 7.6111222244448905e-06, "loss": 0.1731, "step": 5979 }, { "epoch": 1.86, "learning_rate": 7.6107221444288855e-06, "loss": 0.2234, "step": 5980 }, { "epoch": 1.86, "learning_rate": 7.610322064412883e-06, "loss": 0.2613, "step": 5981 }, { "epoch": 1.86, "learning_rate": 7.60992198439688e-06, "loss": 0.1175, "step": 5982 }, { "epoch": 1.86, "learning_rate": 7.609521904380876e-06, "loss": 0.1409, "step": 5983 }, { "epoch": 1.86, "learning_rate": 7.609121824364873e-06, "loss": 0.0572, "step": 5984 }, { "epoch": 1.86, "learning_rate": 7.608721744348871e-06, "loss": 0.1935, "step": 5985 }, { "epoch": 1.86, "learning_rate": 7.608321664332867e-06, "loss": 0.1227, "step": 5986 }, { "epoch": 1.86, "learning_rate": 7.607921584316864e-06, "loss": 0.2424, "step": 5987 }, { "epoch": 1.86, "learning_rate": 7.607521504300861e-06, "loss": 0.1181, "step": 5988 }, { "epoch": 1.86, "learning_rate": 7.607121424284858e-06, "loss": 0.1346, "step": 5989 }, { "epoch": 1.86, "learning_rate": 7.606721344268855e-06, "loss": 0.218, "step": 5990 }, { "epoch": 1.86, "learning_rate": 7.606321264252852e-06, "loss": 0.1503, "step": 5991 }, { "epoch": 1.86, "learning_rate": 7.6059211842368474e-06, "loss": 0.1564, "step": 5992 }, { "epoch": 1.86, "learning_rate": 7.605521104220844e-06, "loss": 0.1046, "step": 5993 }, { "epoch": 1.86, "learning_rate": 7.605121024204841e-06, "loss": 0.1537, "step": 5994 }, { "epoch": 1.86, "learning_rate": 7.604720944188838e-06, "loss": 0.1965, "step": 5995 }, { "epoch": 1.86, "learning_rate": 7.604320864172835e-06, "loss": 0.0782, "step": 5996 }, { "epoch": 1.86, "learning_rate": 7.603920784156832e-06, "loss": 0.2854, "step": 5997 }, { "epoch": 1.86, "learning_rate": 7.603520704140828e-06, "loss": 0.1443, "step": 5998 }, { "epoch": 1.86, "learning_rate": 7.603120624124826e-06, "loss": 0.155, "step": 5999 }, { "epoch": 1.86, "learning_rate": 7.602720544108823e-06, "loss": 0.1252, "step": 6000 }, { "epoch": 1.86, "learning_rate": 7.602320464092819e-06, "loss": 0.0826, "step": 6001 }, { "epoch": 1.86, "learning_rate": 7.601920384076816e-06, "loss": 0.2986, "step": 6002 }, { "epoch": 1.86, "learning_rate": 7.6015203040608135e-06, "loss": 0.2844, "step": 6003 }, { "epoch": 1.86, "learning_rate": 7.60112022404481e-06, "loss": 0.2141, "step": 6004 }, { "epoch": 1.86, "learning_rate": 7.600720144028806e-06, "loss": 0.1948, "step": 6005 }, { "epoch": 1.86, "learning_rate": 7.600320064012803e-06, "loss": 0.1865, "step": 6006 }, { "epoch": 1.86, "learning_rate": 7.599919983996799e-06, "loss": 0.141, "step": 6007 }, { "epoch": 1.86, "learning_rate": 7.599519903980797e-06, "loss": 0.1246, "step": 6008 }, { "epoch": 1.87, "learning_rate": 7.599119823964794e-06, "loss": 0.2082, "step": 6009 }, { "epoch": 1.87, "learning_rate": 7.59871974394879e-06, "loss": 0.1443, "step": 6010 }, { "epoch": 1.87, "learning_rate": 7.598319663932787e-06, "loss": 0.1886, "step": 6011 }, { "epoch": 1.87, "learning_rate": 7.597919583916784e-06, "loss": 0.0863, "step": 6012 }, { "epoch": 1.87, "learning_rate": 7.597519503900781e-06, "loss": 0.1435, "step": 6013 }, { "epoch": 1.87, "learning_rate": 7.597119423884778e-06, "loss": 0.2251, "step": 6014 }, { "epoch": 1.87, "learning_rate": 7.596719343868775e-06, "loss": 0.1299, "step": 6015 }, { "epoch": 1.87, "learning_rate": 7.596319263852771e-06, "loss": 0.1768, "step": 6016 }, { "epoch": 1.87, "learning_rate": 7.595919183836769e-06, "loss": 0.3024, "step": 6017 }, { "epoch": 1.87, "learning_rate": 7.595519103820765e-06, "loss": 0.1726, "step": 6018 }, { "epoch": 1.87, "learning_rate": 7.595119023804761e-06, "loss": 0.1259, "step": 6019 }, { "epoch": 1.87, "learning_rate": 7.594718943788758e-06, "loss": 0.0928, "step": 6020 }, { "epoch": 1.87, "learning_rate": 7.594318863772755e-06, "loss": 0.2018, "step": 6021 }, { "epoch": 1.87, "learning_rate": 7.593918783756752e-06, "loss": 0.062, "step": 6022 }, { "epoch": 1.87, "learning_rate": 7.593518703740749e-06, "loss": 0.083, "step": 6023 }, { "epoch": 1.87, "learning_rate": 7.593118623724746e-06, "loss": 0.2297, "step": 6024 }, { "epoch": 1.87, "learning_rate": 7.592718543708742e-06, "loss": 0.1197, "step": 6025 }, { "epoch": 1.87, "learning_rate": 7.592318463692739e-06, "loss": 0.1797, "step": 6026 }, { "epoch": 1.87, "learning_rate": 7.5919183836767365e-06, "loss": 0.1519, "step": 6027 }, { "epoch": 1.87, "learning_rate": 7.591518303660733e-06, "loss": 0.173, "step": 6028 }, { "epoch": 1.87, "learning_rate": 7.59111822364473e-06, "loss": 0.1372, "step": 6029 }, { "epoch": 1.87, "learning_rate": 7.590718143628726e-06, "loss": 0.1935, "step": 6030 }, { "epoch": 1.87, "learning_rate": 7.5903180636127224e-06, "loss": 0.275, "step": 6031 }, { "epoch": 1.87, "learning_rate": 7.58991798359672e-06, "loss": 0.0948, "step": 6032 }, { "epoch": 1.87, "learning_rate": 7.589517903580717e-06, "loss": 0.0553, "step": 6033 }, { "epoch": 1.87, "learning_rate": 7.589117823564713e-06, "loss": 0.2011, "step": 6034 }, { "epoch": 1.87, "learning_rate": 7.58871774354871e-06, "loss": 0.1644, "step": 6035 }, { "epoch": 1.87, "learning_rate": 7.5883176635327076e-06, "loss": 0.1343, "step": 6036 }, { "epoch": 1.87, "learning_rate": 7.587917583516704e-06, "loss": 0.2074, "step": 6037 }, { "epoch": 1.87, "learning_rate": 7.587517503500701e-06, "loss": 0.1048, "step": 6038 }, { "epoch": 1.87, "learning_rate": 7.587117423484698e-06, "loss": 0.1013, "step": 6039 }, { "epoch": 1.87, "learning_rate": 7.586717343468694e-06, "loss": 0.1148, "step": 6040 }, { "epoch": 1.87, "learning_rate": 7.586317263452692e-06, "loss": 0.1724, "step": 6041 }, { "epoch": 1.88, "learning_rate": 7.5859171834366885e-06, "loss": 0.1106, "step": 6042 }, { "epoch": 1.88, "learning_rate": 7.585517103420684e-06, "loss": 0.1735, "step": 6043 }, { "epoch": 1.88, "learning_rate": 7.585117023404681e-06, "loss": 0.1367, "step": 6044 }, { "epoch": 1.88, "learning_rate": 7.584716943388678e-06, "loss": 0.0679, "step": 6045 }, { "epoch": 1.88, "learning_rate": 7.584316863372675e-06, "loss": 0.1282, "step": 6046 }, { "epoch": 1.88, "learning_rate": 7.583916783356672e-06, "loss": 0.1136, "step": 6047 }, { "epoch": 1.88, "learning_rate": 7.583516703340669e-06, "loss": 0.124, "step": 6048 }, { "epoch": 1.88, "learning_rate": 7.583116623324665e-06, "loss": 0.0899, "step": 6049 }, { "epoch": 1.88, "learning_rate": 7.582716543308663e-06, "loss": 0.1149, "step": 6050 }, { "epoch": 1.88, "learning_rate": 7.5823164632926595e-06, "loss": 0.1784, "step": 6051 }, { "epoch": 1.88, "learning_rate": 7.581916383276656e-06, "loss": 0.099, "step": 6052 }, { "epoch": 1.88, "learning_rate": 7.581516303260653e-06, "loss": 0.1049, "step": 6053 }, { "epoch": 1.88, "learning_rate": 7.58111622324465e-06, "loss": 0.256, "step": 6054 }, { "epoch": 1.88, "learning_rate": 7.5807161432286454e-06, "loss": 0.1637, "step": 6055 }, { "epoch": 1.88, "learning_rate": 7.580316063212643e-06, "loss": 0.2098, "step": 6056 }, { "epoch": 1.88, "learning_rate": 7.57991598319664e-06, "loss": 0.2067, "step": 6057 }, { "epoch": 1.88, "learning_rate": 7.579515903180636e-06, "loss": 0.2255, "step": 6058 }, { "epoch": 1.88, "learning_rate": 7.579115823164633e-06, "loss": 0.1395, "step": 6059 }, { "epoch": 1.88, "learning_rate": 7.5787157431486306e-06, "loss": 0.1201, "step": 6060 }, { "epoch": 1.88, "learning_rate": 7.578315663132627e-06, "loss": 0.111, "step": 6061 }, { "epoch": 1.88, "learning_rate": 7.577915583116624e-06, "loss": 0.1385, "step": 6062 }, { "epoch": 1.88, "learning_rate": 7.577515503100621e-06, "loss": 0.1601, "step": 6063 }, { "epoch": 1.88, "learning_rate": 7.577115423084618e-06, "loss": 0.1321, "step": 6064 }, { "epoch": 1.88, "learning_rate": 7.576715343068615e-06, "loss": 0.2843, "step": 6065 }, { "epoch": 1.88, "learning_rate": 7.5763152630526115e-06, "loss": 0.1915, "step": 6066 }, { "epoch": 1.88, "learning_rate": 7.5763152630526115e-06, "loss": 0.2874, "step": 6067 }, { "epoch": 1.88, "learning_rate": 7.575915183036608e-06, "loss": 0.1584, "step": 6068 }, { "epoch": 1.88, "learning_rate": 7.575515103020604e-06, "loss": 0.1553, "step": 6069 }, { "epoch": 1.88, "learning_rate": 7.575115023004601e-06, "loss": 0.2792, "step": 6070 }, { "epoch": 1.88, "learning_rate": 7.574714942988598e-06, "loss": 0.1699, "step": 6071 }, { "epoch": 1.88, "learning_rate": 7.574314862972595e-06, "loss": 0.1045, "step": 6072 }, { "epoch": 1.88, "learning_rate": 7.573914782956592e-06, "loss": 0.3066, "step": 6073 }, { "epoch": 1.89, "learning_rate": 7.573514702940588e-06, "loss": 0.1439, "step": 6074 }, { "epoch": 1.89, "learning_rate": 7.573114622924586e-06, "loss": 0.1286, "step": 6075 }, { "epoch": 1.89, "learning_rate": 7.5727145429085826e-06, "loss": 0.0828, "step": 6076 }, { "epoch": 1.89, "learning_rate": 7.572314462892579e-06, "loss": 0.2165, "step": 6077 }, { "epoch": 1.89, "learning_rate": 7.571914382876576e-06, "loss": 0.0759, "step": 6078 }, { "epoch": 1.89, "learning_rate": 7.5715143028605735e-06, "loss": 0.3008, "step": 6079 }, { "epoch": 1.89, "learning_rate": 7.57111422284457e-06, "loss": 0.1535, "step": 6080 }, { "epoch": 1.89, "learning_rate": 7.570714142828567e-06, "loss": 0.2431, "step": 6081 }, { "epoch": 1.89, "learning_rate": 7.570314062812563e-06, "loss": 0.1107, "step": 6082 }, { "epoch": 1.89, "learning_rate": 7.569913982796559e-06, "loss": 0.2488, "step": 6083 }, { "epoch": 1.89, "learning_rate": 7.569513902780556e-06, "loss": 0.205, "step": 6084 }, { "epoch": 1.89, "learning_rate": 7.569113822764554e-06, "loss": 0.1109, "step": 6085 }, { "epoch": 1.89, "learning_rate": 7.56871374274855e-06, "loss": 0.1011, "step": 6086 }, { "epoch": 1.89, "learning_rate": 7.568313662732547e-06, "loss": 0.0766, "step": 6087 }, { "epoch": 1.89, "learning_rate": 7.567913582716544e-06, "loss": 0.185, "step": 6088 }, { "epoch": 1.89, "learning_rate": 7.567513502700541e-06, "loss": 0.1965, "step": 6089 }, { "epoch": 1.89, "learning_rate": 7.567113422684538e-06, "loss": 0.2725, "step": 6090 }, { "epoch": 1.89, "learning_rate": 7.5667133426685345e-06, "loss": 0.3033, "step": 6091 }, { "epoch": 1.89, "learning_rate": 7.566313262652531e-06, "loss": 0.2005, "step": 6092 }, { "epoch": 1.89, "learning_rate": 7.565913182636529e-06, "loss": 0.1112, "step": 6093 }, { "epoch": 1.89, "learning_rate": 7.565513102620525e-06, "loss": 0.2561, "step": 6094 }, { "epoch": 1.89, "learning_rate": 7.565113022604521e-06, "loss": 0.1799, "step": 6095 }, { "epoch": 1.89, "learning_rate": 7.564712942588518e-06, "loss": 0.1503, "step": 6096 }, { "epoch": 1.89, "learning_rate": 7.564312862572515e-06, "loss": 0.1776, "step": 6097 }, { "epoch": 1.89, "learning_rate": 7.563912782556511e-06, "loss": 0.1115, "step": 6098 }, { "epoch": 1.89, "learning_rate": 7.563512702540509e-06, "loss": 0.1419, "step": 6099 }, { "epoch": 1.89, "learning_rate": 7.5631126225245056e-06, "loss": 0.0749, "step": 6100 }, { "epoch": 1.89, "learning_rate": 7.562712542508502e-06, "loss": 0.2228, "step": 6101 }, { "epoch": 1.89, "learning_rate": 7.562312462492499e-06, "loss": 0.0877, "step": 6102 }, { "epoch": 1.89, "learning_rate": 7.5619123824764965e-06, "loss": 0.0998, "step": 6103 }, { "epoch": 1.89, "learning_rate": 7.561512302460493e-06, "loss": 0.1034, "step": 6104 }, { "epoch": 1.89, "learning_rate": 7.56111222244449e-06, "loss": 0.0917, "step": 6105 }, { "epoch": 1.9, "learning_rate": 7.5607121424284865e-06, "loss": 0.1676, "step": 6106 }, { "epoch": 1.9, "learning_rate": 7.560312062412482e-06, "loss": 0.1466, "step": 6107 }, { "epoch": 1.9, "learning_rate": 7.55991198239648e-06, "loss": 0.1787, "step": 6108 }, { "epoch": 1.9, "learning_rate": 7.559511902380477e-06, "loss": 0.1471, "step": 6109 }, { "epoch": 1.9, "learning_rate": 7.559111822364473e-06, "loss": 0.1279, "step": 6110 }, { "epoch": 1.9, "learning_rate": 7.55871174234847e-06, "loss": 0.1312, "step": 6111 }, { "epoch": 1.9, "learning_rate": 7.558311662332467e-06, "loss": 0.2286, "step": 6112 }, { "epoch": 1.9, "learning_rate": 7.557911582316464e-06, "loss": 0.1145, "step": 6113 }, { "epoch": 1.9, "learning_rate": 7.557511502300461e-06, "loss": 0.2471, "step": 6114 }, { "epoch": 1.9, "learning_rate": 7.5571114222844576e-06, "loss": 0.159, "step": 6115 }, { "epoch": 1.9, "learning_rate": 7.556711342268454e-06, "loss": 0.2446, "step": 6116 }, { "epoch": 1.9, "learning_rate": 7.556311262252452e-06, "loss": 0.0609, "step": 6117 }, { "epoch": 1.9, "learning_rate": 7.5559111822364485e-06, "loss": 0.1385, "step": 6118 }, { "epoch": 1.9, "learning_rate": 7.555511102220445e-06, "loss": 0.1858, "step": 6119 }, { "epoch": 1.9, "learning_rate": 7.555111022204441e-06, "loss": 0.2507, "step": 6120 }, { "epoch": 1.9, "learning_rate": 7.554710942188438e-06, "loss": 0.1982, "step": 6121 }, { "epoch": 1.9, "learning_rate": 7.554310862172435e-06, "loss": 0.1437, "step": 6122 }, { "epoch": 1.9, "learning_rate": 7.553910782156432e-06, "loss": 0.177, "step": 6123 }, { "epoch": 1.9, "learning_rate": 7.553510702140429e-06, "loss": 0.1404, "step": 6124 }, { "epoch": 1.9, "learning_rate": 7.553110622124425e-06, "loss": 0.167, "step": 6125 }, { "epoch": 1.9, "learning_rate": 7.552710542108422e-06, "loss": 0.149, "step": 6126 }, { "epoch": 1.9, "learning_rate": 7.5523104620924195e-06, "loss": 0.1254, "step": 6127 }, { "epoch": 1.9, "learning_rate": 7.551910382076416e-06, "loss": 0.1553, "step": 6128 }, { "epoch": 1.9, "learning_rate": 7.551510302060413e-06, "loss": 0.119, "step": 6129 }, { "epoch": 1.9, "learning_rate": 7.5511102220444095e-06, "loss": 0.1607, "step": 6130 }, { "epoch": 1.9, "learning_rate": 7.550710142028407e-06, "loss": 0.1215, "step": 6131 }, { "epoch": 1.9, "learning_rate": 7.550310062012403e-06, "loss": 0.1155, "step": 6132 }, { "epoch": 1.9, "learning_rate": 7.5499099819964e-06, "loss": 0.1862, "step": 6133 }, { "epoch": 1.9, "learning_rate": 7.549509901980396e-06, "loss": 0.1097, "step": 6134 }, { "epoch": 1.9, "learning_rate": 7.549109821964393e-06, "loss": 0.0814, "step": 6135 }, { "epoch": 1.9, "learning_rate": 7.5487097419483905e-06, "loss": 0.1935, "step": 6136 }, { "epoch": 1.9, "learning_rate": 7.548309661932387e-06, "loss": 0.0924, "step": 6137 }, { "epoch": 1.91, "learning_rate": 7.547909581916384e-06, "loss": 0.1318, "step": 6138 }, { "epoch": 1.91, "learning_rate": 7.5475095019003806e-06, "loss": 0.0975, "step": 6139 }, { "epoch": 1.91, "learning_rate": 7.547109421884377e-06, "loss": 0.1044, "step": 6140 }, { "epoch": 1.91, "learning_rate": 7.546709341868375e-06, "loss": 0.1365, "step": 6141 }, { "epoch": 1.91, "learning_rate": 7.5463092618523715e-06, "loss": 0.1076, "step": 6142 }, { "epoch": 1.91, "learning_rate": 7.545909181836368e-06, "loss": 0.1567, "step": 6143 }, { "epoch": 1.91, "learning_rate": 7.545509101820365e-06, "loss": 0.0855, "step": 6144 }, { "epoch": 1.91, "learning_rate": 7.545109021804361e-06, "loss": 0.0865, "step": 6145 }, { "epoch": 1.91, "learning_rate": 7.544708941788358e-06, "loss": 0.0602, "step": 6146 }, { "epoch": 1.91, "learning_rate": 7.544308861772355e-06, "loss": 0.0671, "step": 6147 }, { "epoch": 1.91, "learning_rate": 7.543908781756352e-06, "loss": 0.1407, "step": 6148 }, { "epoch": 1.91, "learning_rate": 7.543508701740348e-06, "loss": 0.0888, "step": 6149 }, { "epoch": 1.91, "learning_rate": 7.543108621724346e-06, "loss": 0.2297, "step": 6150 }, { "epoch": 1.91, "learning_rate": 7.5427085417083425e-06, "loss": 0.2694, "step": 6151 }, { "epoch": 1.91, "learning_rate": 7.542308461692339e-06, "loss": 0.1724, "step": 6152 }, { "epoch": 1.91, "learning_rate": 7.541908381676336e-06, "loss": 0.1728, "step": 6153 }, { "epoch": 1.91, "learning_rate": 7.541508301660333e-06, "loss": 0.126, "step": 6154 }, { "epoch": 1.91, "learning_rate": 7.54110822164433e-06, "loss": 0.1503, "step": 6155 }, { "epoch": 1.91, "learning_rate": 7.540708141628327e-06, "loss": 0.172, "step": 6156 }, { "epoch": 1.91, "learning_rate": 7.540308061612323e-06, "loss": 0.2991, "step": 6157 }, { "epoch": 1.91, "learning_rate": 7.539907981596319e-06, "loss": 0.0959, "step": 6158 }, { "epoch": 1.91, "learning_rate": 7.539507901580316e-06, "loss": 0.2283, "step": 6159 }, { "epoch": 1.91, "learning_rate": 7.5391078215643135e-06, "loss": 0.2171, "step": 6160 }, { "epoch": 1.91, "learning_rate": 7.53870774154831e-06, "loss": 0.1083, "step": 6161 }, { "epoch": 1.91, "learning_rate": 7.538307661532307e-06, "loss": 0.1841, "step": 6162 }, { "epoch": 1.91, "learning_rate": 7.537907581516304e-06, "loss": 0.1314, "step": 6163 }, { "epoch": 1.91, "learning_rate": 7.537507501500301e-06, "loss": 0.1753, "step": 6164 }, { "epoch": 1.91, "learning_rate": 7.537107421484298e-06, "loss": 0.159, "step": 6165 }, { "epoch": 1.91, "learning_rate": 7.5367073414682945e-06, "loss": 0.1285, "step": 6166 }, { "epoch": 1.91, "learning_rate": 7.536307261452291e-06, "loss": 0.1447, "step": 6167 }, { "epoch": 1.91, "learning_rate": 7.535907181436289e-06, "loss": 0.1072, "step": 6168 }, { "epoch": 1.91, "learning_rate": 7.535507101420285e-06, "loss": 0.1684, "step": 6169 }, { "epoch": 1.92, "learning_rate": 7.535107021404281e-06, "loss": 0.0752, "step": 6170 }, { "epoch": 1.92, "learning_rate": 7.534706941388278e-06, "loss": 0.2531, "step": 6171 }, { "epoch": 1.92, "learning_rate": 7.534306861372275e-06, "loss": 0.1785, "step": 6172 }, { "epoch": 1.92, "learning_rate": 7.533906781356271e-06, "loss": 0.1616, "step": 6173 }, { "epoch": 1.92, "learning_rate": 7.533506701340269e-06, "loss": 0.0711, "step": 6174 }, { "epoch": 1.92, "learning_rate": 7.5331066213242655e-06, "loss": 0.1145, "step": 6175 }, { "epoch": 1.92, "learning_rate": 7.532706541308262e-06, "loss": 0.1558, "step": 6176 }, { "epoch": 1.92, "learning_rate": 7.532306461292259e-06, "loss": 0.1481, "step": 6177 }, { "epoch": 1.92, "learning_rate": 7.531906381276256e-06, "loss": 0.1293, "step": 6178 }, { "epoch": 1.92, "learning_rate": 7.531506301260253e-06, "loss": 0.0655, "step": 6179 }, { "epoch": 1.92, "learning_rate": 7.53110622124425e-06, "loss": 0.11, "step": 6180 }, { "epoch": 1.92, "learning_rate": 7.5307061412282465e-06, "loss": 0.1513, "step": 6181 }, { "epoch": 1.92, "learning_rate": 7.530306061212244e-06, "loss": 0.095, "step": 6182 }, { "epoch": 1.92, "learning_rate": 7.529905981196239e-06, "loss": 0.1874, "step": 6183 }, { "epoch": 1.92, "learning_rate": 7.5295059011802365e-06, "loss": 0.1565, "step": 6184 }, { "epoch": 1.92, "learning_rate": 7.529105821164233e-06, "loss": 0.0875, "step": 6185 }, { "epoch": 1.92, "learning_rate": 7.52870574114823e-06, "loss": 0.2209, "step": 6186 }, { "epoch": 1.92, "learning_rate": 7.528305661132227e-06, "loss": 0.2729, "step": 6187 }, { "epoch": 1.92, "learning_rate": 7.527905581116224e-06, "loss": 0.1815, "step": 6188 }, { "epoch": 1.92, "learning_rate": 7.527505501100221e-06, "loss": 0.0949, "step": 6189 }, { "epoch": 1.92, "learning_rate": 7.5271054210842175e-06, "loss": 0.1129, "step": 6190 }, { "epoch": 1.92, "learning_rate": 7.526705341068214e-06, "loss": 0.1409, "step": 6191 }, { "epoch": 1.92, "learning_rate": 7.526305261052212e-06, "loss": 0.2653, "step": 6192 }, { "epoch": 1.92, "learning_rate": 7.525905181036208e-06, "loss": 0.1532, "step": 6193 }, { "epoch": 1.92, "learning_rate": 7.525505101020205e-06, "loss": 0.1643, "step": 6194 }, { "epoch": 1.92, "learning_rate": 7.525105021004201e-06, "loss": 0.1205, "step": 6195 }, { "epoch": 1.92, "learning_rate": 7.524704940988198e-06, "loss": 0.097, "step": 6196 }, { "epoch": 1.92, "learning_rate": 7.524304860972194e-06, "loss": 0.1632, "step": 6197 }, { "epoch": 1.92, "learning_rate": 7.523904780956192e-06, "loss": 0.1006, "step": 6198 }, { "epoch": 1.92, "learning_rate": 7.5235047009401885e-06, "loss": 0.1475, "step": 6199 }, { "epoch": 1.92, "learning_rate": 7.523104620924185e-06, "loss": 0.1062, "step": 6200 }, { "epoch": 1.92, "learning_rate": 7.522704540908182e-06, "loss": 0.0925, "step": 6201 }, { "epoch": 1.92, "learning_rate": 7.5223044608921794e-06, "loss": 0.177, "step": 6202 }, { "epoch": 1.93, "learning_rate": 7.521904380876176e-06, "loss": 0.0882, "step": 6203 }, { "epoch": 1.93, "learning_rate": 7.521504300860173e-06, "loss": 0.2048, "step": 6204 }, { "epoch": 1.93, "learning_rate": 7.5211042208441695e-06, "loss": 0.144, "step": 6205 }, { "epoch": 1.93, "learning_rate": 7.520704140828167e-06, "loss": 0.1478, "step": 6206 }, { "epoch": 1.93, "learning_rate": 7.520304060812164e-06, "loss": 0.1327, "step": 6207 }, { "epoch": 1.93, "learning_rate": 7.5199039807961595e-06, "loss": 0.1176, "step": 6208 }, { "epoch": 1.93, "learning_rate": 7.519503900780156e-06, "loss": 0.2458, "step": 6209 }, { "epoch": 1.93, "learning_rate": 7.519103820764153e-06, "loss": 0.112, "step": 6210 }, { "epoch": 1.93, "learning_rate": 7.51870374074815e-06, "loss": 0.2692, "step": 6211 }, { "epoch": 1.93, "learning_rate": 7.518303660732147e-06, "loss": 0.1522, "step": 6212 }, { "epoch": 1.93, "learning_rate": 7.517903580716144e-06, "loss": 0.1928, "step": 6213 }, { "epoch": 1.93, "learning_rate": 7.5175035007001405e-06, "loss": 0.1766, "step": 6214 }, { "epoch": 1.93, "learning_rate": 7.517103420684137e-06, "loss": 0.1236, "step": 6215 }, { "epoch": 1.93, "learning_rate": 7.516703340668135e-06, "loss": 0.104, "step": 6216 }, { "epoch": 1.93, "learning_rate": 7.516303260652131e-06, "loss": 0.1366, "step": 6217 }, { "epoch": 1.93, "learning_rate": 7.515903180636128e-06, "loss": 0.1337, "step": 6218 }, { "epoch": 1.93, "learning_rate": 7.515503100620125e-06, "loss": 0.2467, "step": 6219 }, { "epoch": 1.93, "learning_rate": 7.515103020604121e-06, "loss": 0.2231, "step": 6220 }, { "epoch": 1.93, "learning_rate": 7.514702940588118e-06, "loss": 0.1633, "step": 6221 }, { "epoch": 1.93, "learning_rate": 7.514302860572115e-06, "loss": 0.1252, "step": 6222 }, { "epoch": 1.93, "learning_rate": 7.5139027805561115e-06, "loss": 0.1495, "step": 6223 }, { "epoch": 1.93, "learning_rate": 7.513502700540108e-06, "loss": 0.1407, "step": 6224 }, { "epoch": 1.93, "learning_rate": 7.513102620524106e-06, "loss": 0.1038, "step": 6225 }, { "epoch": 1.93, "learning_rate": 7.5127025405081024e-06, "loss": 0.306, "step": 6226 }, { "epoch": 1.93, "learning_rate": 7.512302460492099e-06, "loss": 0.1006, "step": 6227 }, { "epoch": 1.93, "learning_rate": 7.511902380476096e-06, "loss": 0.0387, "step": 6228 }, { "epoch": 1.93, "learning_rate": 7.5115023004600925e-06, "loss": 0.1163, "step": 6229 }, { "epoch": 1.93, "learning_rate": 7.51110222044409e-06, "loss": 0.1887, "step": 6230 }, { "epoch": 1.93, "learning_rate": 7.510702140428087e-06, "loss": 0.1271, "step": 6231 }, { "epoch": 1.93, "learning_rate": 7.510302060412083e-06, "loss": 0.1553, "step": 6232 }, { "epoch": 1.93, "learning_rate": 7.509901980396079e-06, "loss": 0.1499, "step": 6233 }, { "epoch": 1.93, "learning_rate": 7.509501900380076e-06, "loss": 0.1131, "step": 6234 }, { "epoch": 1.94, "learning_rate": 7.5091018203640735e-06, "loss": 0.1858, "step": 6235 }, { "epoch": 1.94, "learning_rate": 7.50870174034807e-06, "loss": 0.2438, "step": 6236 }, { "epoch": 1.94, "learning_rate": 7.508301660332067e-06, "loss": 0.1457, "step": 6237 }, { "epoch": 1.94, "learning_rate": 7.5079015803160635e-06, "loss": 0.101, "step": 6238 }, { "epoch": 1.94, "learning_rate": 7.507501500300061e-06, "loss": 0.1959, "step": 6239 }, { "epoch": 1.94, "learning_rate": 7.507101420284058e-06, "loss": 0.1598, "step": 6240 }, { "epoch": 1.94, "learning_rate": 7.5067013402680544e-06, "loss": 0.1112, "step": 6241 }, { "epoch": 1.94, "learning_rate": 7.506301260252051e-06, "loss": 0.0975, "step": 6242 }, { "epoch": 1.94, "learning_rate": 7.505901180236048e-06, "loss": 0.1514, "step": 6243 }, { "epoch": 1.94, "learning_rate": 7.505501100220045e-06, "loss": 0.1292, "step": 6244 }, { "epoch": 1.94, "learning_rate": 7.505101020204042e-06, "loss": 0.1161, "step": 6245 }, { "epoch": 1.94, "learning_rate": 7.504700940188038e-06, "loss": 0.249, "step": 6246 }, { "epoch": 1.94, "learning_rate": 7.5043008601720345e-06, "loss": 0.108, "step": 6247 }, { "epoch": 1.94, "learning_rate": 7.503900780156031e-06, "loss": 0.2661, "step": 6248 }, { "epoch": 1.94, "learning_rate": 7.503500700140029e-06, "loss": 0.1773, "step": 6249 }, { "epoch": 1.94, "learning_rate": 7.5031006201240254e-06, "loss": 0.2744, "step": 6250 }, { "epoch": 1.94, "learning_rate": 7.502700540108022e-06, "loss": 0.1034, "step": 6251 }, { "epoch": 1.94, "learning_rate": 7.502300460092019e-06, "loss": 0.1689, "step": 6252 }, { "epoch": 1.94, "learning_rate": 7.501900380076016e-06, "loss": 0.0727, "step": 6253 }, { "epoch": 1.94, "learning_rate": 7.501500300060013e-06, "loss": 0.1257, "step": 6254 }, { "epoch": 1.94, "learning_rate": 7.50110022004401e-06, "loss": 0.1376, "step": 6255 }, { "epoch": 1.94, "learning_rate": 7.500700140028006e-06, "loss": 0.1864, "step": 6256 }, { "epoch": 1.94, "learning_rate": 7.500300060012003e-06, "loss": 0.1501, "step": 6257 }, { "epoch": 1.94, "learning_rate": 7.499899979995999e-06, "loss": 0.194, "step": 6258 }, { "epoch": 1.94, "learning_rate": 7.4994998999799965e-06, "loss": 0.1753, "step": 6259 }, { "epoch": 1.94, "learning_rate": 7.499099819963993e-06, "loss": 0.1683, "step": 6260 }, { "epoch": 1.94, "learning_rate": 7.49869973994799e-06, "loss": 0.1353, "step": 6261 }, { "epoch": 1.94, "learning_rate": 7.4982996599319865e-06, "loss": 0.1109, "step": 6262 }, { "epoch": 1.94, "learning_rate": 7.497899579915984e-06, "loss": 0.1013, "step": 6263 }, { "epoch": 1.94, "learning_rate": 7.497499499899981e-06, "loss": 0.1534, "step": 6264 }, { "epoch": 1.94, "learning_rate": 7.4970994198839774e-06, "loss": 0.1236, "step": 6265 }, { "epoch": 1.94, "learning_rate": 7.496699339867974e-06, "loss": 0.1196, "step": 6266 }, { "epoch": 1.95, "learning_rate": 7.496299259851972e-06, "loss": 0.0673, "step": 6267 }, { "epoch": 1.95, "learning_rate": 7.495899179835968e-06, "loss": 0.1532, "step": 6268 }, { "epoch": 1.95, "learning_rate": 7.495499099819965e-06, "loss": 0.1484, "step": 6269 }, { "epoch": 1.95, "learning_rate": 7.495099019803962e-06, "loss": 0.1171, "step": 6270 }, { "epoch": 1.95, "learning_rate": 7.4946989397879576e-06, "loss": 0.136, "step": 6271 }, { "epoch": 1.95, "learning_rate": 7.494298859771954e-06, "loss": 0.1623, "step": 6272 }, { "epoch": 1.95, "learning_rate": 7.493898779755952e-06, "loss": 0.2074, "step": 6273 }, { "epoch": 1.95, "learning_rate": 7.4934986997399485e-06, "loss": 0.0955, "step": 6274 }, { "epoch": 1.95, "learning_rate": 7.493098619723945e-06, "loss": 0.1211, "step": 6275 }, { "epoch": 1.95, "learning_rate": 7.492698539707942e-06, "loss": 0.1399, "step": 6276 }, { "epoch": 1.95, "learning_rate": 7.492298459691939e-06, "loss": 0.1169, "step": 6277 }, { "epoch": 1.95, "learning_rate": 7.491898379675936e-06, "loss": 0.1039, "step": 6278 }, { "epoch": 1.95, "learning_rate": 7.491498299659933e-06, "loss": 0.1637, "step": 6279 }, { "epoch": 1.95, "learning_rate": 7.4910982196439294e-06, "loss": 0.2348, "step": 6280 }, { "epoch": 1.95, "learning_rate": 7.490698139627927e-06, "loss": 0.1867, "step": 6281 }, { "epoch": 1.95, "learning_rate": 7.490298059611924e-06, "loss": 0.1147, "step": 6282 }, { "epoch": 1.95, "learning_rate": 7.4898979795959195e-06, "loss": 0.0584, "step": 6283 }, { "epoch": 1.95, "learning_rate": 7.489497899579916e-06, "loss": 0.0579, "step": 6284 }, { "epoch": 1.95, "learning_rate": 7.489097819563913e-06, "loss": 0.1412, "step": 6285 }, { "epoch": 1.95, "learning_rate": 7.4886977395479095e-06, "loss": 0.0952, "step": 6286 }, { "epoch": 1.95, "learning_rate": 7.488297659531907e-06, "loss": 0.1751, "step": 6287 }, { "epoch": 1.95, "learning_rate": 7.487897579515904e-06, "loss": 0.093, "step": 6288 }, { "epoch": 1.95, "learning_rate": 7.4874974994999004e-06, "loss": 0.2128, "step": 6289 }, { "epoch": 1.95, "learning_rate": 7.487097419483897e-06, "loss": 0.2064, "step": 6290 }, { "epoch": 1.95, "learning_rate": 7.486697339467895e-06, "loss": 0.0789, "step": 6291 }, { "epoch": 1.95, "learning_rate": 7.486297259451891e-06, "loss": 0.0834, "step": 6292 }, { "epoch": 1.95, "learning_rate": 7.485897179435888e-06, "loss": 0.1166, "step": 6293 }, { "epoch": 1.95, "learning_rate": 7.485497099419885e-06, "loss": 0.1289, "step": 6294 }, { "epoch": 1.95, "learning_rate": 7.485097019403882e-06, "loss": 0.2303, "step": 6295 }, { "epoch": 1.95, "learning_rate": 7.484696939387877e-06, "loss": 0.2001, "step": 6296 }, { "epoch": 1.95, "learning_rate": 7.484296859371875e-06, "loss": 0.151, "step": 6297 }, { "epoch": 1.95, "learning_rate": 7.4838967793558715e-06, "loss": 0.2412, "step": 6298 }, { "epoch": 1.96, "learning_rate": 7.483496699339868e-06, "loss": 0.1712, "step": 6299 }, { "epoch": 1.96, "learning_rate": 7.483096619323865e-06, "loss": 0.1368, "step": 6300 }, { "epoch": 1.96, "learning_rate": 7.482696539307862e-06, "loss": 0.0884, "step": 6301 }, { "epoch": 1.96, "learning_rate": 7.482296459291859e-06, "loss": 0.1841, "step": 6302 }, { "epoch": 1.96, "learning_rate": 7.481896379275856e-06, "loss": 0.1803, "step": 6303 }, { "epoch": 1.96, "learning_rate": 7.4814962992598524e-06, "loss": 0.1267, "step": 6304 }, { "epoch": 1.96, "learning_rate": 7.48109621924385e-06, "loss": 0.1835, "step": 6305 }, { "epoch": 1.96, "learning_rate": 7.480696139227847e-06, "loss": 0.1088, "step": 6306 }, { "epoch": 1.96, "learning_rate": 7.480296059211843e-06, "loss": 0.1201, "step": 6307 }, { "epoch": 1.96, "learning_rate": 7.47989597919584e-06, "loss": 0.1262, "step": 6308 }, { "epoch": 1.96, "learning_rate": 7.479495899179836e-06, "loss": 0.217, "step": 6309 }, { "epoch": 1.96, "learning_rate": 7.479095819163833e-06, "loss": 0.0895, "step": 6310 }, { "epoch": 1.96, "learning_rate": 7.47869573914783e-06, "loss": 0.1308, "step": 6311 }, { "epoch": 1.96, "learning_rate": 7.478295659131827e-06, "loss": 0.1117, "step": 6312 }, { "epoch": 1.96, "learning_rate": 7.4778955791158235e-06, "loss": 0.1023, "step": 6313 }, { "epoch": 1.96, "learning_rate": 7.47749549909982e-06, "loss": 0.16, "step": 6314 }, { "epoch": 1.96, "learning_rate": 7.477095419083818e-06, "loss": 0.1942, "step": 6315 }, { "epoch": 1.96, "learning_rate": 7.476695339067814e-06, "loss": 0.1404, "step": 6316 }, { "epoch": 1.96, "learning_rate": 7.476295259051811e-06, "loss": 0.1504, "step": 6317 }, { "epoch": 1.96, "learning_rate": 7.475895179035808e-06, "loss": 0.1205, "step": 6318 }, { "epoch": 1.96, "learning_rate": 7.475495099019805e-06, "loss": 0.1012, "step": 6319 }, { "epoch": 1.96, "learning_rate": 7.475095019003802e-06, "loss": 0.0966, "step": 6320 }, { "epoch": 1.96, "learning_rate": 7.474694938987798e-06, "loss": 0.105, "step": 6321 }, { "epoch": 1.96, "learning_rate": 7.4742948589717945e-06, "loss": 0.1603, "step": 6322 }, { "epoch": 1.96, "learning_rate": 7.473894778955791e-06, "loss": 0.1855, "step": 6323 }, { "epoch": 1.96, "learning_rate": 7.473494698939789e-06, "loss": 0.1162, "step": 6324 }, { "epoch": 1.96, "learning_rate": 7.473094618923785e-06, "loss": 0.1253, "step": 6325 }, { "epoch": 1.96, "learning_rate": 7.472694538907782e-06, "loss": 0.2043, "step": 6326 }, { "epoch": 1.96, "learning_rate": 7.472294458891779e-06, "loss": 0.2455, "step": 6327 }, { "epoch": 1.96, "learning_rate": 7.4718943788757754e-06, "loss": 0.1024, "step": 6328 }, { "epoch": 1.96, "learning_rate": 7.471494298859773e-06, "loss": 0.1704, "step": 6329 }, { "epoch": 1.96, "learning_rate": 7.47109421884377e-06, "loss": 0.1446, "step": 6330 }, { "epoch": 1.97, "learning_rate": 7.470694138827766e-06, "loss": 0.3063, "step": 6331 }, { "epoch": 1.97, "learning_rate": 7.470294058811763e-06, "loss": 0.3344, "step": 6332 }, { "epoch": 1.97, "learning_rate": 7.4698939787957606e-06, "loss": 0.09, "step": 6333 }, { "epoch": 1.97, "learning_rate": 7.469493898779756e-06, "loss": 0.2535, "step": 6334 }, { "epoch": 1.97, "learning_rate": 7.469093818763753e-06, "loss": 0.1814, "step": 6335 }, { "epoch": 1.97, "learning_rate": 7.46869373874775e-06, "loss": 0.258, "step": 6336 }, { "epoch": 1.97, "learning_rate": 7.4682936587317465e-06, "loss": 0.213, "step": 6337 }, { "epoch": 1.97, "learning_rate": 7.467893578715744e-06, "loss": 0.1423, "step": 6338 }, { "epoch": 1.97, "learning_rate": 7.467493498699741e-06, "loss": 0.1239, "step": 6339 }, { "epoch": 1.97, "learning_rate": 7.467093418683737e-06, "loss": 0.1264, "step": 6340 }, { "epoch": 1.97, "learning_rate": 7.466693338667734e-06, "loss": 0.2625, "step": 6341 }, { "epoch": 1.97, "learning_rate": 7.466293258651731e-06, "loss": 0.0941, "step": 6342 }, { "epoch": 1.97, "learning_rate": 7.465893178635728e-06, "loss": 0.2779, "step": 6343 }, { "epoch": 1.97, "learning_rate": 7.465493098619725e-06, "loss": 0.1354, "step": 6344 }, { "epoch": 1.97, "learning_rate": 7.465093018603722e-06, "loss": 0.1347, "step": 6345 }, { "epoch": 1.97, "learning_rate": 7.464692938587718e-06, "loss": 0.1687, "step": 6346 }, { "epoch": 1.97, "learning_rate": 7.464292858571714e-06, "loss": 0.2185, "step": 6347 }, { "epoch": 1.97, "learning_rate": 7.463892778555712e-06, "loss": 0.1874, "step": 6348 }, { "epoch": 1.97, "learning_rate": 7.463492698539708e-06, "loss": 0.0679, "step": 6349 }, { "epoch": 1.97, "learning_rate": 7.463092618523705e-06, "loss": 0.1135, "step": 6350 }, { "epoch": 1.97, "learning_rate": 7.462692538507702e-06, "loss": 0.1671, "step": 6351 }, { "epoch": 1.97, "learning_rate": 7.462292458491699e-06, "loss": 0.1338, "step": 6352 }, { "epoch": 1.97, "learning_rate": 7.461892378475696e-06, "loss": 0.1459, "step": 6353 }, { "epoch": 1.97, "learning_rate": 7.461492298459693e-06, "loss": 0.1523, "step": 6354 }, { "epoch": 1.97, "learning_rate": 7.461092218443689e-06, "loss": 0.23, "step": 6355 }, { "epoch": 1.97, "learning_rate": 7.460692138427686e-06, "loss": 0.1169, "step": 6356 }, { "epoch": 1.97, "learning_rate": 7.460292058411684e-06, "loss": 0.0973, "step": 6357 }, { "epoch": 1.97, "learning_rate": 7.45989197839568e-06, "loss": 0.0827, "step": 6358 }, { "epoch": 1.97, "learning_rate": 7.459491898379676e-06, "loss": 0.0687, "step": 6359 }, { "epoch": 1.97, "learning_rate": 7.459091818363673e-06, "loss": 0.121, "step": 6360 }, { "epoch": 1.97, "learning_rate": 7.4586917383476695e-06, "loss": 0.0847, "step": 6361 }, { "epoch": 1.97, "learning_rate": 7.458291658331667e-06, "loss": 0.113, "step": 6362 }, { "epoch": 1.97, "learning_rate": 7.457891578315664e-06, "loss": 0.1424, "step": 6363 }, { "epoch": 1.98, "learning_rate": 7.45749149829966e-06, "loss": 0.1777, "step": 6364 }, { "epoch": 1.98, "learning_rate": 7.457091418283657e-06, "loss": 0.1016, "step": 6365 }, { "epoch": 1.98, "learning_rate": 7.456691338267655e-06, "loss": 0.1359, "step": 6366 }, { "epoch": 1.98, "learning_rate": 7.456291258251651e-06, "loss": 0.1291, "step": 6367 }, { "epoch": 1.98, "learning_rate": 7.455891178235648e-06, "loss": 0.1555, "step": 6368 }, { "epoch": 1.98, "learning_rate": 7.455491098219645e-06, "loss": 0.1233, "step": 6369 }, { "epoch": 1.98, "learning_rate": 7.455091018203642e-06, "loss": 0.2069, "step": 6370 }, { "epoch": 1.98, "learning_rate": 7.454690938187639e-06, "loss": 0.0757, "step": 6371 }, { "epoch": 1.98, "learning_rate": 7.454290858171635e-06, "loss": 0.289, "step": 6372 }, { "epoch": 1.98, "learning_rate": 7.453890778155631e-06, "loss": 0.1244, "step": 6373 }, { "epoch": 1.98, "learning_rate": 7.453490698139628e-06, "loss": 0.1055, "step": 6374 }, { "epoch": 1.98, "learning_rate": 7.453090618123625e-06, "loss": 0.1351, "step": 6375 }, { "epoch": 1.98, "learning_rate": 7.452690538107622e-06, "loss": 0.0953, "step": 6376 }, { "epoch": 1.98, "learning_rate": 7.452290458091619e-06, "loss": 0.1926, "step": 6377 }, { "epoch": 1.98, "learning_rate": 7.451890378075616e-06, "loss": 0.1485, "step": 6378 }, { "epoch": 1.98, "learning_rate": 7.451490298059612e-06, "loss": 0.1614, "step": 6379 }, { "epoch": 1.98, "learning_rate": 7.45109021804361e-06, "loss": 0.1988, "step": 6380 }, { "epoch": 1.98, "learning_rate": 7.450690138027607e-06, "loss": 0.1243, "step": 6381 }, { "epoch": 1.98, "learning_rate": 7.450290058011603e-06, "loss": 0.2165, "step": 6382 }, { "epoch": 1.98, "learning_rate": 7.4498899779956e-06, "loss": 0.3059, "step": 6383 }, { "epoch": 1.98, "learning_rate": 7.449489897979596e-06, "loss": 0.1472, "step": 6384 }, { "epoch": 1.98, "learning_rate": 7.4490898179635925e-06, "loss": 0.1814, "step": 6385 }, { "epoch": 1.98, "learning_rate": 7.44868973794759e-06, "loss": 0.0803, "step": 6386 }, { "epoch": 1.98, "learning_rate": 7.448289657931587e-06, "loss": 0.2218, "step": 6387 }, { "epoch": 1.98, "learning_rate": 7.447889577915583e-06, "loss": 0.1002, "step": 6388 }, { "epoch": 1.98, "learning_rate": 7.44748949789958e-06, "loss": 0.155, "step": 6389 }, { "epoch": 1.98, "learning_rate": 7.447089417883578e-06, "loss": 0.0982, "step": 6390 }, { "epoch": 1.98, "learning_rate": 7.446689337867574e-06, "loss": 0.1407, "step": 6391 }, { "epoch": 1.98, "learning_rate": 7.446289257851571e-06, "loss": 0.0988, "step": 6392 }, { "epoch": 1.98, "learning_rate": 7.445889177835568e-06, "loss": 0.0975, "step": 6393 }, { "epoch": 1.98, "learning_rate": 7.445489097819565e-06, "loss": 0.1002, "step": 6394 }, { "epoch": 1.98, "learning_rate": 7.445089017803562e-06, "loss": 0.3297, "step": 6395 }, { "epoch": 1.99, "learning_rate": 7.444688937787559e-06, "loss": 0.1527, "step": 6396 }, { "epoch": 1.99, "learning_rate": 7.4442888577715544e-06, "loss": 0.0863, "step": 6397 }, { "epoch": 1.99, "learning_rate": 7.443888777755551e-06, "loss": 0.1594, "step": 6398 }, { "epoch": 1.99, "learning_rate": 7.443488697739548e-06, "loss": 0.2599, "step": 6399 }, { "epoch": 1.99, "learning_rate": 7.443088617723545e-06, "loss": 0.224, "step": 6400 }, { "epoch": 1.99, "learning_rate": 7.442688537707542e-06, "loss": 0.1146, "step": 6401 }, { "epoch": 1.99, "learning_rate": 7.442288457691539e-06, "loss": 0.166, "step": 6402 }, { "epoch": 1.99, "learning_rate": 7.441888377675535e-06, "loss": 0.1302, "step": 6403 }, { "epoch": 1.99, "learning_rate": 7.441488297659533e-06, "loss": 0.1257, "step": 6404 }, { "epoch": 1.99, "learning_rate": 7.44108821764353e-06, "loss": 0.2422, "step": 6405 }, { "epoch": 1.99, "learning_rate": 7.440688137627526e-06, "loss": 0.1496, "step": 6406 }, { "epoch": 1.99, "learning_rate": 7.440288057611523e-06, "loss": 0.1156, "step": 6407 }, { "epoch": 1.99, "learning_rate": 7.4398879775955205e-06, "loss": 0.1359, "step": 6408 }, { "epoch": 1.99, "learning_rate": 7.439487897579517e-06, "loss": 0.1868, "step": 6409 }, { "epoch": 1.99, "learning_rate": 7.439087817563513e-06, "loss": 0.1337, "step": 6410 }, { "epoch": 1.99, "learning_rate": 7.43868773754751e-06, "loss": 0.1726, "step": 6411 }, { "epoch": 1.99, "learning_rate": 7.438287657531506e-06, "loss": 0.2166, "step": 6412 }, { "epoch": 1.99, "learning_rate": 7.437887577515503e-06, "loss": 0.067, "step": 6413 }, { "epoch": 1.99, "learning_rate": 7.437487497499501e-06, "loss": 0.1831, "step": 6414 }, { "epoch": 1.99, "learning_rate": 7.437087417483497e-06, "loss": 0.2458, "step": 6415 }, { "epoch": 1.99, "learning_rate": 7.436687337467494e-06, "loss": 0.1545, "step": 6416 }, { "epoch": 1.99, "learning_rate": 7.436287257451491e-06, "loss": 0.1431, "step": 6417 }, { "epoch": 1.99, "learning_rate": 7.435887177435488e-06, "loss": 0.1329, "step": 6418 }, { "epoch": 1.99, "learning_rate": 7.435487097419485e-06, "loss": 0.1359, "step": 6419 }, { "epoch": 1.99, "learning_rate": 7.435087017403482e-06, "loss": 0.1739, "step": 6420 }, { "epoch": 1.99, "learning_rate": 7.434686937387478e-06, "loss": 0.1432, "step": 6421 }, { "epoch": 1.99, "learning_rate": 7.434286857371474e-06, "loss": 0.1547, "step": 6422 }, { "epoch": 1.99, "learning_rate": 7.433886777355472e-06, "loss": 0.1242, "step": 6423 }, { "epoch": 1.99, "learning_rate": 7.433486697339468e-06, "loss": 0.1573, "step": 6424 }, { "epoch": 1.99, "learning_rate": 7.433086617323465e-06, "loss": 0.2066, "step": 6425 }, { "epoch": 1.99, "learning_rate": 7.432686537307462e-06, "loss": 0.2017, "step": 6426 }, { "epoch": 1.99, "learning_rate": 7.432286457291458e-06, "loss": 0.1201, "step": 6427 }, { "epoch": 2.0, "learning_rate": 7.431886377275456e-06, "loss": 0.1233, "step": 6428 }, { "epoch": 2.0, "learning_rate": 7.431486297259453e-06, "loss": 0.3157, "step": 6429 }, { "epoch": 2.0, "learning_rate": 7.431086217243449e-06, "loss": 0.0831, "step": 6430 }, { "epoch": 2.0, "learning_rate": 7.430686137227446e-06, "loss": 0.3103, "step": 6431 }, { "epoch": 2.0, "learning_rate": 7.4302860572114435e-06, "loss": 0.1286, "step": 6432 }, { "epoch": 2.0, "learning_rate": 7.42988597719544e-06, "loss": 0.1138, "step": 6433 }, { "epoch": 2.0, "learning_rate": 7.429485897179437e-06, "loss": 0.1609, "step": 6434 }, { "epoch": 2.0, "learning_rate": 7.429085817163433e-06, "loss": 0.147, "step": 6435 }, { "epoch": 2.0, "learning_rate": 7.4286857371474294e-06, "loss": 0.2036, "step": 6436 }, { "epoch": 2.0, "learning_rate": 7.428285657131427e-06, "loss": 0.1063, "step": 6437 }, { "epoch": 2.0, "learning_rate": 7.427885577115424e-06, "loss": 0.126, "step": 6438 }, { "epoch": 2.0, "learning_rate": 7.42748549709942e-06, "loss": 0.1048, "step": 6439 }, { "epoch": 2.0, "learning_rate": 7.427085417083417e-06, "loss": 0.0722, "step": 6440 }, { "epoch": 2.0, "learning_rate": 7.426685337067414e-06, "loss": 0.0709, "step": 6441 }, { "epoch": 2.0, "learning_rate": 7.426285257051411e-06, "loss": 0.0838, "step": 6442 }, { "epoch": 2.0, "learning_rate": 7.425885177035408e-06, "loss": 0.1922, "step": 6443 }, { "epoch": 2.0, "learning_rate": 7.425485097019405e-06, "loss": 0.0836, "step": 6444 }, { "epoch": 2.0, "learning_rate": 7.425085017003401e-06, "loss": 0.1248, "step": 6445 }, { "epoch": 2.0, "learning_rate": 7.424684936987399e-06, "loss": 0.1099, "step": 6446 }, { "epoch": 2.0, "learning_rate": 7.424284856971395e-06, "loss": 0.0755, "step": 6447 }, { "epoch": 2.0, "learning_rate": 7.423884776955391e-06, "loss": 0.0817, "step": 6448 }, { "epoch": 2.0, "learning_rate": 7.423484696939388e-06, "loss": 0.093, "step": 6449 }, { "epoch": 2.0, "learning_rate": 7.423084616923385e-06, "loss": 0.1144, "step": 6450 }, { "epoch": 2.0, "learning_rate": 7.422684536907382e-06, "loss": 0.0633, "step": 6451 }, { "epoch": 2.0, "learning_rate": 7.422284456891379e-06, "loss": 0.0731, "step": 6452 }, { "epoch": 2.0, "learning_rate": 7.421884376875376e-06, "loss": 0.0996, "step": 6453 }, { "epoch": 2.0, "learning_rate": 7.421484296859372e-06, "loss": 0.0974, "step": 6454 }, { "epoch": 2.0, "learning_rate": 7.42108421684337e-06, "loss": 0.0359, "step": 6455 }, { "epoch": 2.0, "learning_rate": 7.4206841368273665e-06, "loss": 0.0596, "step": 6456 }, { "epoch": 2.0, "learning_rate": 7.420284056811363e-06, "loss": 0.0595, "step": 6457 }, { "epoch": 2.0, "learning_rate": 7.41988397679536e-06, "loss": 0.057, "step": 6458 }, { "epoch": 2.0, "learning_rate": 7.419483896779357e-06, "loss": 0.1065, "step": 6459 }, { "epoch": 2.01, "learning_rate": 7.4190838167633524e-06, "loss": 0.0743, "step": 6460 }, { "epoch": 2.01, "learning_rate": 7.41868373674735e-06, "loss": 0.0567, "step": 6461 }, { "epoch": 2.01, "learning_rate": 7.418283656731347e-06, "loss": 0.1147, "step": 6462 }, { "epoch": 2.01, "learning_rate": 7.417883576715343e-06, "loss": 0.0838, "step": 6463 }, { "epoch": 2.01, "learning_rate": 7.41748349669934e-06, "loss": 0.1229, "step": 6464 }, { "epoch": 2.01, "learning_rate": 7.4170834166833376e-06, "loss": 0.1014, "step": 6465 }, { "epoch": 2.01, "learning_rate": 7.416683336667334e-06, "loss": 0.1005, "step": 6466 }, { "epoch": 2.01, "learning_rate": 7.416283256651331e-06, "loss": 0.0619, "step": 6467 }, { "epoch": 2.01, "learning_rate": 7.415883176635328e-06, "loss": 0.043, "step": 6468 }, { "epoch": 2.01, "learning_rate": 7.415483096619325e-06, "loss": 0.0634, "step": 6469 }, { "epoch": 2.01, "learning_rate": 7.415083016603322e-06, "loss": 0.1703, "step": 6470 }, { "epoch": 2.01, "learning_rate": 7.4146829365873185e-06, "loss": 0.0749, "step": 6471 }, { "epoch": 2.01, "learning_rate": 7.414282856571315e-06, "loss": 0.1442, "step": 6472 }, { "epoch": 2.01, "learning_rate": 7.413882776555311e-06, "loss": 0.0744, "step": 6473 }, { "epoch": 2.01, "learning_rate": 7.413482696539308e-06, "loss": 0.0641, "step": 6474 }, { "epoch": 2.01, "learning_rate": 7.413082616523305e-06, "loss": 0.0535, "step": 6475 }, { "epoch": 2.01, "learning_rate": 7.412682536507302e-06, "loss": 0.1029, "step": 6476 }, { "epoch": 2.01, "learning_rate": 7.412282456491299e-06, "loss": 0.0387, "step": 6477 }, { "epoch": 2.01, "learning_rate": 7.411882376475295e-06, "loss": 0.1101, "step": 6478 }, { "epoch": 2.01, "learning_rate": 7.411482296459293e-06, "loss": 0.0848, "step": 6479 }, { "epoch": 2.01, "learning_rate": 7.4110822164432895e-06, "loss": 0.1564, "step": 6480 }, { "epoch": 2.01, "learning_rate": 7.410682136427286e-06, "loss": 0.0855, "step": 6481 }, { "epoch": 2.01, "learning_rate": 7.410282056411283e-06, "loss": 0.048, "step": 6482 }, { "epoch": 2.01, "learning_rate": 7.4098819763952805e-06, "loss": 0.0694, "step": 6483 }, { "epoch": 2.01, "learning_rate": 7.409481896379277e-06, "loss": 0.0516, "step": 6484 }, { "epoch": 2.01, "learning_rate": 7.409081816363273e-06, "loss": 0.0402, "step": 6485 }, { "epoch": 2.01, "learning_rate": 7.40868173634727e-06, "loss": 0.191, "step": 6486 }, { "epoch": 2.01, "learning_rate": 7.408281656331266e-06, "loss": 0.0894, "step": 6487 }, { "epoch": 2.01, "learning_rate": 7.407881576315263e-06, "loss": 0.0901, "step": 6488 }, { "epoch": 2.01, "learning_rate": 7.4074814962992606e-06, "loss": 0.127, "step": 6489 }, { "epoch": 2.01, "learning_rate": 7.407081416283257e-06, "loss": 0.0723, "step": 6490 }, { "epoch": 2.01, "learning_rate": 7.406681336267254e-06, "loss": 0.0262, "step": 6491 }, { "epoch": 2.01, "learning_rate": 7.406281256251251e-06, "loss": 0.1687, "step": 6492 }, { "epoch": 2.02, "learning_rate": 7.405881176235248e-06, "loss": 0.0785, "step": 6493 }, { "epoch": 2.02, "learning_rate": 7.405481096219245e-06, "loss": 0.0744, "step": 6494 }, { "epoch": 2.02, "learning_rate": 7.4050810162032415e-06, "loss": 0.1058, "step": 6495 }, { "epoch": 2.02, "learning_rate": 7.404680936187238e-06, "loss": 0.0661, "step": 6496 }, { "epoch": 2.02, "learning_rate": 7.404280856171236e-06, "loss": 0.0629, "step": 6497 }, { "epoch": 2.02, "learning_rate": 7.403880776155231e-06, "loss": 0.0459, "step": 6498 }, { "epoch": 2.02, "learning_rate": 7.403480696139228e-06, "loss": 0.0925, "step": 6499 }, { "epoch": 2.02, "learning_rate": 7.403080616123225e-06, "loss": 0.0611, "step": 6500 }, { "epoch": 2.02, "learning_rate": 7.402680536107222e-06, "loss": 0.084, "step": 6501 }, { "epoch": 2.02, "learning_rate": 7.402280456091218e-06, "loss": 0.0988, "step": 6502 }, { "epoch": 2.02, "learning_rate": 7.401880376075216e-06, "loss": 0.1579, "step": 6503 }, { "epoch": 2.02, "learning_rate": 7.4014802960592126e-06, "loss": 0.0545, "step": 6504 }, { "epoch": 2.02, "learning_rate": 7.401080216043209e-06, "loss": 0.1039, "step": 6505 }, { "epoch": 2.02, "learning_rate": 7.400680136027206e-06, "loss": 0.1319, "step": 6506 }, { "epoch": 2.02, "learning_rate": 7.4002800560112035e-06, "loss": 0.0567, "step": 6507 }, { "epoch": 2.02, "learning_rate": 7.3998799759952e-06, "loss": 0.0714, "step": 6508 }, { "epoch": 2.02, "learning_rate": 7.399479895979197e-06, "loss": 0.1089, "step": 6509 }, { "epoch": 2.02, "learning_rate": 7.399079815963193e-06, "loss": 0.0716, "step": 6510 }, { "epoch": 2.02, "learning_rate": 7.398679735947189e-06, "loss": 0.1126, "step": 6511 }, { "epoch": 2.02, "learning_rate": 7.398279655931186e-06, "loss": 0.1139, "step": 6512 }, { "epoch": 2.02, "learning_rate": 7.397879575915184e-06, "loss": 0.145, "step": 6513 }, { "epoch": 2.02, "learning_rate": 7.39747949589918e-06, "loss": 0.0584, "step": 6514 }, { "epoch": 2.02, "learning_rate": 7.397079415883177e-06, "loss": 0.0503, "step": 6515 }, { "epoch": 2.02, "learning_rate": 7.396679335867174e-06, "loss": 0.1032, "step": 6516 }, { "epoch": 2.02, "learning_rate": 7.396279255851171e-06, "loss": 0.1131, "step": 6517 }, { "epoch": 2.02, "learning_rate": 7.395879175835168e-06, "loss": 0.0489, "step": 6518 }, { "epoch": 2.02, "learning_rate": 7.3954790958191645e-06, "loss": 0.076, "step": 6519 }, { "epoch": 2.02, "learning_rate": 7.395079015803161e-06, "loss": 0.1544, "step": 6520 }, { "epoch": 2.02, "learning_rate": 7.394678935787159e-06, "loss": 0.2012, "step": 6521 }, { "epoch": 2.02, "learning_rate": 7.3942788557711555e-06, "loss": 0.0451, "step": 6522 }, { "epoch": 2.02, "learning_rate": 7.393878775755151e-06, "loss": 0.1358, "step": 6523 }, { "epoch": 2.02, "learning_rate": 7.393478695739148e-06, "loss": 0.0733, "step": 6524 }, { "epoch": 2.03, "learning_rate": 7.393078615723145e-06, "loss": 0.153, "step": 6525 }, { "epoch": 2.03, "learning_rate": 7.392678535707142e-06, "loss": 0.0926, "step": 6526 }, { "epoch": 2.03, "learning_rate": 7.392278455691139e-06, "loss": 0.0603, "step": 6527 }, { "epoch": 2.03, "learning_rate": 7.3918783756751356e-06, "loss": 0.1587, "step": 6528 }, { "epoch": 2.03, "learning_rate": 7.391478295659132e-06, "loss": 0.0441, "step": 6529 }, { "epoch": 2.03, "learning_rate": 7.391078215643129e-06, "loss": 0.0642, "step": 6530 }, { "epoch": 2.03, "learning_rate": 7.3906781356271265e-06, "loss": 0.0991, "step": 6531 }, { "epoch": 2.03, "learning_rate": 7.390278055611123e-06, "loss": 0.1058, "step": 6532 }, { "epoch": 2.03, "learning_rate": 7.38987797559512e-06, "loss": 0.0494, "step": 6533 }, { "epoch": 2.03, "learning_rate": 7.3894778955791165e-06, "loss": 0.0829, "step": 6534 }, { "epoch": 2.03, "learning_rate": 7.389077815563114e-06, "loss": 0.0769, "step": 6535 }, { "epoch": 2.03, "learning_rate": 7.38867773554711e-06, "loss": 0.0283, "step": 6536 }, { "epoch": 2.03, "learning_rate": 7.388277655531107e-06, "loss": 0.0384, "step": 6537 }, { "epoch": 2.03, "learning_rate": 7.387877575515103e-06, "loss": 0.0664, "step": 6538 }, { "epoch": 2.03, "learning_rate": 7.3874774954991e-06, "loss": 0.0825, "step": 6539 }, { "epoch": 2.03, "learning_rate": 7.3870774154830975e-06, "loss": 0.0729, "step": 6540 }, { "epoch": 2.03, "learning_rate": 7.386677335467094e-06, "loss": 0.0799, "step": 6541 }, { "epoch": 2.03, "learning_rate": 7.386277255451091e-06, "loss": 0.0632, "step": 6542 }, { "epoch": 2.03, "learning_rate": 7.3858771754350876e-06, "loss": 0.1177, "step": 6543 }, { "epoch": 2.03, "learning_rate": 7.385477095419084e-06, "loss": 0.0779, "step": 6544 }, { "epoch": 2.03, "learning_rate": 7.385077015403082e-06, "loss": 0.0641, "step": 6545 }, { "epoch": 2.03, "learning_rate": 7.3846769353870785e-06, "loss": 0.0793, "step": 6546 }, { "epoch": 2.03, "learning_rate": 7.384276855371075e-06, "loss": 0.1749, "step": 6547 }, { "epoch": 2.03, "learning_rate": 7.383876775355071e-06, "loss": 0.0567, "step": 6548 }, { "epoch": 2.03, "learning_rate": 7.383476695339068e-06, "loss": 0.0605, "step": 6549 }, { "epoch": 2.03, "learning_rate": 7.383076615323065e-06, "loss": 0.0782, "step": 6550 }, { "epoch": 2.03, "learning_rate": 7.382676535307062e-06, "loss": 0.0438, "step": 6551 }, { "epoch": 2.03, "learning_rate": 7.382276455291059e-06, "loss": 0.0647, "step": 6552 }, { "epoch": 2.03, "learning_rate": 7.381876375275055e-06, "loss": 0.0937, "step": 6553 }, { "epoch": 2.03, "learning_rate": 7.381476295259053e-06, "loss": 0.1904, "step": 6554 }, { "epoch": 2.03, "learning_rate": 7.3810762152430495e-06, "loss": 0.069, "step": 6555 }, { "epoch": 2.03, "learning_rate": 7.380676135227046e-06, "loss": 0.0583, "step": 6556 }, { "epoch": 2.04, "learning_rate": 7.380276055211043e-06, "loss": 0.0709, "step": 6557 }, { "epoch": 2.04, "learning_rate": 7.3798759751950395e-06, "loss": 0.0712, "step": 6558 }, { "epoch": 2.04, "learning_rate": 7.379475895179037e-06, "loss": 0.0576, "step": 6559 }, { "epoch": 2.04, "learning_rate": 7.379075815163034e-06, "loss": 0.083, "step": 6560 }, { "epoch": 2.04, "learning_rate": 7.37867573514703e-06, "loss": 0.0461, "step": 6561 }, { "epoch": 2.04, "learning_rate": 7.378275655131026e-06, "loss": 0.1666, "step": 6562 }, { "epoch": 2.04, "learning_rate": 7.377875575115023e-06, "loss": 0.0896, "step": 6563 }, { "epoch": 2.04, "learning_rate": 7.3774754950990205e-06, "loss": 0.1182, "step": 6564 }, { "epoch": 2.04, "learning_rate": 7.377075415083017e-06, "loss": 0.0693, "step": 6565 }, { "epoch": 2.04, "learning_rate": 7.376675335067014e-06, "loss": 0.0439, "step": 6566 }, { "epoch": 2.04, "learning_rate": 7.3762752550510106e-06, "loss": 0.036, "step": 6567 }, { "epoch": 2.04, "learning_rate": 7.375875175035008e-06, "loss": 0.0978, "step": 6568 }, { "epoch": 2.04, "learning_rate": 7.375475095019005e-06, "loss": 0.1999, "step": 6569 }, { "epoch": 2.04, "learning_rate": 7.3750750150030015e-06, "loss": 0.2091, "step": 6570 }, { "epoch": 2.04, "learning_rate": 7.374674934986998e-06, "loss": 0.0878, "step": 6571 }, { "epoch": 2.04, "learning_rate": 7.374274854970995e-06, "loss": 0.1161, "step": 6572 }, { "epoch": 2.04, "learning_rate": 7.373874774954992e-06, "loss": 0.073, "step": 6573 }, { "epoch": 2.04, "learning_rate": 7.373474694938988e-06, "loss": 0.0887, "step": 6574 }, { "epoch": 2.04, "learning_rate": 7.373074614922985e-06, "loss": 0.0363, "step": 6575 }, { "epoch": 2.04, "learning_rate": 7.372674534906982e-06, "loss": 0.0455, "step": 6576 }, { "epoch": 2.04, "learning_rate": 7.372274454890978e-06, "loss": 0.0488, "step": 6577 }, { "epoch": 2.04, "learning_rate": 7.371874374874976e-06, "loss": 0.2329, "step": 6578 }, { "epoch": 2.04, "learning_rate": 7.3714742948589725e-06, "loss": 0.0408, "step": 6579 }, { "epoch": 2.04, "learning_rate": 7.371074214842969e-06, "loss": 0.1106, "step": 6580 }, { "epoch": 2.04, "learning_rate": 7.370674134826966e-06, "loss": 0.1106, "step": 6581 }, { "epoch": 2.04, "learning_rate": 7.370274054810963e-06, "loss": 0.0622, "step": 6582 }, { "epoch": 2.04, "learning_rate": 7.36987397479496e-06, "loss": 0.0871, "step": 6583 }, { "epoch": 2.04, "learning_rate": 7.369473894778957e-06, "loss": 0.0803, "step": 6584 }, { "epoch": 2.04, "learning_rate": 7.3690738147629535e-06, "loss": 0.0798, "step": 6585 }, { "epoch": 2.04, "learning_rate": 7.368673734746949e-06, "loss": 0.0263, "step": 6586 }, { "epoch": 2.04, "learning_rate": 7.368273654730946e-06, "loss": 0.0775, "step": 6587 }, { "epoch": 2.04, "learning_rate": 7.3678735747149435e-06, "loss": 0.0363, "step": 6588 }, { "epoch": 2.05, "learning_rate": 7.36747349469894e-06, "loss": 0.1354, "step": 6589 }, { "epoch": 2.05, "learning_rate": 7.367073414682937e-06, "loss": 0.1455, "step": 6590 }, { "epoch": 2.05, "learning_rate": 7.366673334666934e-06, "loss": 0.0917, "step": 6591 }, { "epoch": 2.05, "learning_rate": 7.366273254650931e-06, "loss": 0.0494, "step": 6592 }, { "epoch": 2.05, "learning_rate": 7.365873174634928e-06, "loss": 0.1936, "step": 6593 }, { "epoch": 2.05, "learning_rate": 7.3654730946189245e-06, "loss": 0.136, "step": 6594 }, { "epoch": 2.05, "learning_rate": 7.365073014602921e-06, "loss": 0.065, "step": 6595 }, { "epoch": 2.05, "learning_rate": 7.364672934586919e-06, "loss": 0.0987, "step": 6596 }, { "epoch": 2.05, "learning_rate": 7.364272854570915e-06, "loss": 0.0792, "step": 6597 }, { "epoch": 2.05, "learning_rate": 7.363872774554912e-06, "loss": 0.0812, "step": 6598 }, { "epoch": 2.05, "learning_rate": 7.363472694538908e-06, "loss": 0.0786, "step": 6599 }, { "epoch": 2.05, "learning_rate": 7.363072614522905e-06, "loss": 0.042, "step": 6600 }, { "epoch": 2.05, "learning_rate": 7.362672534506901e-06, "loss": 0.0697, "step": 6601 }, { "epoch": 2.05, "learning_rate": 7.362272454490899e-06, "loss": 0.0678, "step": 6602 }, { "epoch": 2.05, "learning_rate": 7.3618723744748955e-06, "loss": 0.1998, "step": 6603 }, { "epoch": 2.05, "learning_rate": 7.361472294458892e-06, "loss": 0.0577, "step": 6604 }, { "epoch": 2.05, "learning_rate": 7.361072214442889e-06, "loss": 0.0655, "step": 6605 }, { "epoch": 2.05, "learning_rate": 7.360672134426886e-06, "loss": 0.0758, "step": 6606 }, { "epoch": 2.05, "learning_rate": 7.360272054410883e-06, "loss": 0.1055, "step": 6607 }, { "epoch": 2.05, "learning_rate": 7.35987197439488e-06, "loss": 0.0931, "step": 6608 }, { "epoch": 2.05, "learning_rate": 7.3594718943788765e-06, "loss": 0.061, "step": 6609 }, { "epoch": 2.05, "learning_rate": 7.359071814362874e-06, "loss": 0.0782, "step": 6610 }, { "epoch": 2.05, "learning_rate": 7.35867173434687e-06, "loss": 0.0924, "step": 6611 }, { "epoch": 2.05, "learning_rate": 7.3582716543308665e-06, "loss": 0.1289, "step": 6612 }, { "epoch": 2.05, "learning_rate": 7.357871574314863e-06, "loss": 0.0587, "step": 6613 }, { "epoch": 2.05, "learning_rate": 7.35747149429886e-06, "loss": 0.0558, "step": 6614 }, { "epoch": 2.05, "learning_rate": 7.357071414282857e-06, "loss": 0.1113, "step": 6615 }, { "epoch": 2.05, "learning_rate": 7.356671334266854e-06, "loss": 0.0782, "step": 6616 }, { "epoch": 2.05, "learning_rate": 7.356271254250851e-06, "loss": 0.0501, "step": 6617 }, { "epoch": 2.05, "learning_rate": 7.3558711742348475e-06, "loss": 0.0934, "step": 6618 }, { "epoch": 2.05, "learning_rate": 7.355471094218844e-06, "loss": 0.1093, "step": 6619 }, { "epoch": 2.05, "learning_rate": 7.355071014202842e-06, "loss": 0.0802, "step": 6620 }, { "epoch": 2.06, "learning_rate": 7.354670934186838e-06, "loss": 0.0318, "step": 6621 }, { "epoch": 2.06, "learning_rate": 7.354270854170835e-06, "loss": 0.1114, "step": 6622 }, { "epoch": 2.06, "learning_rate": 7.353870774154832e-06, "loss": 0.1884, "step": 6623 }, { "epoch": 2.06, "learning_rate": 7.353470694138828e-06, "loss": 0.0996, "step": 6624 }, { "epoch": 2.06, "learning_rate": 7.353070614122825e-06, "loss": 0.0394, "step": 6625 }, { "epoch": 2.06, "learning_rate": 7.352670534106822e-06, "loss": 0.1108, "step": 6626 }, { "epoch": 2.06, "learning_rate": 7.3522704540908185e-06, "loss": 0.0593, "step": 6627 }, { "epoch": 2.06, "learning_rate": 7.351870374074815e-06, "loss": 0.0968, "step": 6628 }, { "epoch": 2.06, "learning_rate": 7.351470294058812e-06, "loss": 0.0451, "step": 6629 }, { "epoch": 2.06, "learning_rate": 7.3510702140428094e-06, "loss": 0.0563, "step": 6630 }, { "epoch": 2.06, "learning_rate": 7.350670134026806e-06, "loss": 0.1928, "step": 6631 }, { "epoch": 2.06, "learning_rate": 7.350270054010803e-06, "loss": 0.1312, "step": 6632 }, { "epoch": 2.06, "learning_rate": 7.3498699739947995e-06, "loss": 0.0709, "step": 6633 }, { "epoch": 2.06, "learning_rate": 7.349469893978797e-06, "loss": 0.0557, "step": 6634 }, { "epoch": 2.06, "learning_rate": 7.349069813962794e-06, "loss": 0.0834, "step": 6635 }, { "epoch": 2.06, "learning_rate": 7.34866973394679e-06, "loss": 0.0831, "step": 6636 }, { "epoch": 2.06, "learning_rate": 7.348269653930786e-06, "loss": 0.1741, "step": 6637 }, { "epoch": 2.06, "learning_rate": 7.347869573914783e-06, "loss": 0.0871, "step": 6638 }, { "epoch": 2.06, "learning_rate": 7.3474694938987805e-06, "loss": 0.0508, "step": 6639 }, { "epoch": 2.06, "learning_rate": 7.347069413882777e-06, "loss": 0.0407, "step": 6640 }, { "epoch": 2.06, "learning_rate": 7.346669333866774e-06, "loss": 0.081, "step": 6641 }, { "epoch": 2.06, "learning_rate": 7.3462692538507705e-06, "loss": 0.074, "step": 6642 }, { "epoch": 2.06, "learning_rate": 7.345869173834767e-06, "loss": 0.0629, "step": 6643 }, { "epoch": 2.06, "learning_rate": 7.345469093818765e-06, "loss": 0.2026, "step": 6644 }, { "epoch": 2.06, "learning_rate": 7.345069013802761e-06, "loss": 0.093, "step": 6645 }, { "epoch": 2.06, "learning_rate": 7.344668933786758e-06, "loss": 0.1388, "step": 6646 }, { "epoch": 2.06, "learning_rate": 7.344268853770755e-06, "loss": 0.1067, "step": 6647 }, { "epoch": 2.06, "learning_rate": 7.343868773754752e-06, "loss": 0.0858, "step": 6648 }, { "epoch": 2.06, "learning_rate": 7.343468693738748e-06, "loss": 0.0807, "step": 6649 }, { "epoch": 2.06, "learning_rate": 7.343068613722745e-06, "loss": 0.0844, "step": 6650 }, { "epoch": 2.06, "learning_rate": 7.3426685337067415e-06, "loss": 0.1506, "step": 6651 }, { "epoch": 2.06, "learning_rate": 7.342268453690738e-06, "loss": 0.0473, "step": 6652 }, { "epoch": 2.06, "learning_rate": 7.341868373674736e-06, "loss": 0.0792, "step": 6653 }, { "epoch": 2.07, "learning_rate": 7.3414682936587324e-06, "loss": 0.1344, "step": 6654 }, { "epoch": 2.07, "learning_rate": 7.341068213642729e-06, "loss": 0.1167, "step": 6655 }, { "epoch": 2.07, "learning_rate": 7.340668133626726e-06, "loss": 0.1003, "step": 6656 }, { "epoch": 2.07, "learning_rate": 7.3402680536107225e-06, "loss": 0.0405, "step": 6657 }, { "epoch": 2.07, "learning_rate": 7.33986797359472e-06, "loss": 0.0721, "step": 6658 }, { "epoch": 2.07, "learning_rate": 7.339467893578717e-06, "loss": 0.1297, "step": 6659 }, { "epoch": 2.07, "learning_rate": 7.339067813562713e-06, "loss": 0.0788, "step": 6660 }, { "epoch": 2.07, "learning_rate": 7.33866773354671e-06, "loss": 0.0857, "step": 6661 }, { "epoch": 2.07, "learning_rate": 7.338267653530706e-06, "loss": 0.1408, "step": 6662 }, { "epoch": 2.07, "learning_rate": 7.3378675735147035e-06, "loss": 0.0646, "step": 6663 }, { "epoch": 2.07, "learning_rate": 7.3374674934987e-06, "loss": 0.0817, "step": 6664 }, { "epoch": 2.07, "learning_rate": 7.337067413482697e-06, "loss": 0.0957, "step": 6665 }, { "epoch": 2.07, "learning_rate": 7.3366673334666935e-06, "loss": 0.0959, "step": 6666 }, { "epoch": 2.07, "learning_rate": 7.336267253450691e-06, "loss": 0.0958, "step": 6667 }, { "epoch": 2.07, "learning_rate": 7.335867173434688e-06, "loss": 0.1338, "step": 6668 }, { "epoch": 2.07, "learning_rate": 7.3354670934186844e-06, "loss": 0.0952, "step": 6669 }, { "epoch": 2.07, "learning_rate": 7.335067013402681e-06, "loss": 0.0432, "step": 6670 }, { "epoch": 2.07, "learning_rate": 7.334666933386679e-06, "loss": 0.0478, "step": 6671 }, { "epoch": 2.07, "learning_rate": 7.334266853370675e-06, "loss": 0.06, "step": 6672 }, { "epoch": 2.07, "learning_rate": 7.333866773354672e-06, "loss": 0.0975, "step": 6673 }, { "epoch": 2.07, "learning_rate": 7.333466693338668e-06, "loss": 0.0882, "step": 6674 }, { "epoch": 2.07, "learning_rate": 7.3330666133226645e-06, "loss": 0.0888, "step": 6675 }, { "epoch": 2.07, "learning_rate": 7.332666533306661e-06, "loss": 0.0335, "step": 6676 }, { "epoch": 2.07, "learning_rate": 7.332266453290659e-06, "loss": 0.106, "step": 6677 }, { "epoch": 2.07, "learning_rate": 7.3318663732746555e-06, "loss": 0.0883, "step": 6678 }, { "epoch": 2.07, "learning_rate": 7.331466293258652e-06, "loss": 0.1941, "step": 6679 }, { "epoch": 2.07, "learning_rate": 7.331066213242649e-06, "loss": 0.0959, "step": 6680 }, { "epoch": 2.07, "learning_rate": 7.330666133226646e-06, "loss": 0.084, "step": 6681 }, { "epoch": 2.07, "learning_rate": 7.330266053210643e-06, "loss": 0.0407, "step": 6682 }, { "epoch": 2.07, "learning_rate": 7.32986597319464e-06, "loss": 0.0897, "step": 6683 }, { "epoch": 2.07, "learning_rate": 7.329465893178636e-06, "loss": 0.0726, "step": 6684 }, { "epoch": 2.07, "learning_rate": 7.329065813162634e-06, "loss": 0.0746, "step": 6685 }, { "epoch": 2.08, "learning_rate": 7.328665733146631e-06, "loss": 0.1115, "step": 6686 }, { "epoch": 2.08, "learning_rate": 7.3282656531306265e-06, "loss": 0.0714, "step": 6687 }, { "epoch": 2.08, "learning_rate": 7.327865573114623e-06, "loss": 0.0447, "step": 6688 }, { "epoch": 2.08, "learning_rate": 7.32746549309862e-06, "loss": 0.0836, "step": 6689 }, { "epoch": 2.08, "learning_rate": 7.3270654130826165e-06, "loss": 0.0583, "step": 6690 }, { "epoch": 2.08, "learning_rate": 7.326665333066614e-06, "loss": 0.1444, "step": 6691 }, { "epoch": 2.08, "learning_rate": 7.326265253050611e-06, "loss": 0.1936, "step": 6692 }, { "epoch": 2.08, "learning_rate": 7.3258651730346074e-06, "loss": 0.0913, "step": 6693 }, { "epoch": 2.08, "learning_rate": 7.325465093018604e-06, "loss": 0.1018, "step": 6694 }, { "epoch": 2.08, "learning_rate": 7.325065013002602e-06, "loss": 0.0618, "step": 6695 }, { "epoch": 2.08, "learning_rate": 7.324664932986598e-06, "loss": 0.0746, "step": 6696 }, { "epoch": 2.08, "learning_rate": 7.324264852970595e-06, "loss": 0.0719, "step": 6697 }, { "epoch": 2.08, "learning_rate": 7.323864772954592e-06, "loss": 0.0697, "step": 6698 }, { "epoch": 2.08, "learning_rate": 7.323464692938589e-06, "loss": 0.1757, "step": 6699 }, { "epoch": 2.08, "learning_rate": 7.323064612922584e-06, "loss": 0.1133, "step": 6700 }, { "epoch": 2.08, "learning_rate": 7.322664532906582e-06, "loss": 0.0616, "step": 6701 }, { "epoch": 2.08, "learning_rate": 7.3222644528905785e-06, "loss": 0.1025, "step": 6702 }, { "epoch": 2.08, "learning_rate": 7.321864372874575e-06, "loss": 0.0743, "step": 6703 }, { "epoch": 2.08, "learning_rate": 7.321464292858572e-06, "loss": 0.2283, "step": 6704 }, { "epoch": 2.08, "learning_rate": 7.321064212842569e-06, "loss": 0.0728, "step": 6705 }, { "epoch": 2.08, "learning_rate": 7.320664132826566e-06, "loss": 0.0684, "step": 6706 }, { "epoch": 2.08, "learning_rate": 7.320264052810563e-06, "loss": 0.0434, "step": 6707 }, { "epoch": 2.08, "learning_rate": 7.3198639727945594e-06, "loss": 0.0814, "step": 6708 }, { "epoch": 2.08, "learning_rate": 7.319463892778557e-06, "loss": 0.0879, "step": 6709 }, { "epoch": 2.08, "learning_rate": 7.319063812762554e-06, "loss": 0.1182, "step": 6710 }, { "epoch": 2.08, "learning_rate": 7.31866373274655e-06, "loss": 0.0538, "step": 6711 }, { "epoch": 2.08, "learning_rate": 7.318263652730546e-06, "loss": 0.0337, "step": 6712 }, { "epoch": 2.08, "learning_rate": 7.317863572714543e-06, "loss": 0.1175, "step": 6713 }, { "epoch": 2.08, "learning_rate": 7.3174634926985395e-06, "loss": 0.0901, "step": 6714 }, { "epoch": 2.08, "learning_rate": 7.317063412682537e-06, "loss": 0.1031, "step": 6715 }, { "epoch": 2.08, "learning_rate": 7.316663332666534e-06, "loss": 0.0576, "step": 6716 }, { "epoch": 2.08, "learning_rate": 7.3162632526505305e-06, "loss": 0.1226, "step": 6717 }, { "epoch": 2.09, "learning_rate": 7.315863172634527e-06, "loss": 0.0957, "step": 6718 }, { "epoch": 2.09, "learning_rate": 7.315463092618525e-06, "loss": 0.0434, "step": 6719 }, { "epoch": 2.09, "learning_rate": 7.315063012602521e-06, "loss": 0.0866, "step": 6720 }, { "epoch": 2.09, "learning_rate": 7.314662932586518e-06, "loss": 0.0792, "step": 6721 }, { "epoch": 2.09, "learning_rate": 7.314262852570515e-06, "loss": 0.079, "step": 6722 }, { "epoch": 2.09, "learning_rate": 7.313862772554512e-06, "loss": 0.0458, "step": 6723 }, { "epoch": 2.09, "learning_rate": 7.313462692538509e-06, "loss": 0.0864, "step": 6724 }, { "epoch": 2.09, "learning_rate": 7.313062612522505e-06, "loss": 0.0658, "step": 6725 }, { "epoch": 2.09, "learning_rate": 7.3126625325065015e-06, "loss": 0.1808, "step": 6726 }, { "epoch": 2.09, "learning_rate": 7.312262452490498e-06, "loss": 0.0382, "step": 6727 }, { "epoch": 2.09, "learning_rate": 7.311862372474495e-06, "loss": 0.0893, "step": 6728 }, { "epoch": 2.09, "learning_rate": 7.311462292458492e-06, "loss": 0.0457, "step": 6729 }, { "epoch": 2.09, "learning_rate": 7.311062212442489e-06, "loss": 0.0907, "step": 6730 }, { "epoch": 2.09, "learning_rate": 7.310662132426486e-06, "loss": 0.1661, "step": 6731 }, { "epoch": 2.09, "learning_rate": 7.3102620524104824e-06, "loss": 0.0587, "step": 6732 }, { "epoch": 2.09, "learning_rate": 7.30986197239448e-06, "loss": 0.1004, "step": 6733 }, { "epoch": 2.09, "learning_rate": 7.309461892378477e-06, "loss": 0.0674, "step": 6734 }, { "epoch": 2.09, "learning_rate": 7.309061812362473e-06, "loss": 0.0829, "step": 6735 }, { "epoch": 2.09, "learning_rate": 7.30866173234647e-06, "loss": 0.0805, "step": 6736 }, { "epoch": 2.09, "learning_rate": 7.308261652330466e-06, "loss": 0.0553, "step": 6737 }, { "epoch": 2.09, "learning_rate": 7.307861572314463e-06, "loss": 0.0797, "step": 6738 }, { "epoch": 2.09, "learning_rate": 7.30746149229846e-06, "loss": 0.081, "step": 6739 }, { "epoch": 2.09, "learning_rate": 7.307061412282457e-06, "loss": 0.1179, "step": 6740 }, { "epoch": 2.09, "learning_rate": 7.3066613322664535e-06, "loss": 0.0769, "step": 6741 }, { "epoch": 2.09, "learning_rate": 7.30626125225045e-06, "loss": 0.0933, "step": 6742 }, { "epoch": 2.09, "learning_rate": 7.305861172234448e-06, "loss": 0.0597, "step": 6743 }, { "epoch": 2.09, "learning_rate": 7.305461092218444e-06, "loss": 0.0394, "step": 6744 }, { "epoch": 2.09, "learning_rate": 7.305061012202441e-06, "loss": 0.0566, "step": 6745 }, { "epoch": 2.09, "learning_rate": 7.304660932186438e-06, "loss": 0.113, "step": 6746 }, { "epoch": 2.09, "learning_rate": 7.304260852170435e-06, "loss": 0.0478, "step": 6747 }, { "epoch": 2.09, "learning_rate": 7.303860772154432e-06, "loss": 0.0839, "step": 6748 }, { "epoch": 2.09, "learning_rate": 7.303460692138429e-06, "loss": 0.0646, "step": 6749 }, { "epoch": 2.1, "learning_rate": 7.3030606121224245e-06, "loss": 0.0858, "step": 6750 }, { "epoch": 2.1, "learning_rate": 7.302660532106421e-06, "loss": 0.0573, "step": 6751 }, { "epoch": 2.1, "learning_rate": 7.302260452090419e-06, "loss": 0.12, "step": 6752 }, { "epoch": 2.1, "learning_rate": 7.301860372074415e-06, "loss": 0.0891, "step": 6753 }, { "epoch": 2.1, "learning_rate": 7.301460292058412e-06, "loss": 0.0719, "step": 6754 }, { "epoch": 2.1, "learning_rate": 7.301060212042409e-06, "loss": 0.1261, "step": 6755 }, { "epoch": 2.1, "learning_rate": 7.300660132026406e-06, "loss": 0.0432, "step": 6756 }, { "epoch": 2.1, "learning_rate": 7.300260052010403e-06, "loss": 0.1495, "step": 6757 }, { "epoch": 2.1, "learning_rate": 7.2998599719944e-06, "loss": 0.0365, "step": 6758 }, { "epoch": 2.1, "learning_rate": 7.299459891978396e-06, "loss": 0.1342, "step": 6759 }, { "epoch": 2.1, "learning_rate": 7.299059811962393e-06, "loss": 0.0921, "step": 6760 }, { "epoch": 2.1, "learning_rate": 7.298659731946391e-06, "loss": 0.1189, "step": 6761 }, { "epoch": 2.1, "learning_rate": 7.298259651930387e-06, "loss": 0.0923, "step": 6762 }, { "epoch": 2.1, "learning_rate": 7.297859571914383e-06, "loss": 0.0833, "step": 6763 }, { "epoch": 2.1, "learning_rate": 7.29745949189838e-06, "loss": 0.0708, "step": 6764 }, { "epoch": 2.1, "learning_rate": 7.2970594118823765e-06, "loss": 0.0806, "step": 6765 }, { "epoch": 2.1, "learning_rate": 7.296659331866374e-06, "loss": 0.1114, "step": 6766 }, { "epoch": 2.1, "learning_rate": 7.296259251850371e-06, "loss": 0.0769, "step": 6767 }, { "epoch": 2.1, "learning_rate": 7.295859171834367e-06, "loss": 0.0666, "step": 6768 }, { "epoch": 2.1, "learning_rate": 7.295459091818364e-06, "loss": 0.0747, "step": 6769 }, { "epoch": 2.1, "learning_rate": 7.295059011802362e-06, "loss": 0.0577, "step": 6770 }, { "epoch": 2.1, "learning_rate": 7.294658931786358e-06, "loss": 0.0456, "step": 6771 }, { "epoch": 2.1, "learning_rate": 7.294258851770355e-06, "loss": 0.0709, "step": 6772 }, { "epoch": 2.1, "learning_rate": 7.293858771754352e-06, "loss": 0.0714, "step": 6773 }, { "epoch": 2.1, "learning_rate": 7.293458691738348e-06, "loss": 0.1247, "step": 6774 }, { "epoch": 2.1, "learning_rate": 7.293058611722344e-06, "loss": 0.0614, "step": 6775 }, { "epoch": 2.1, "learning_rate": 7.292658531706342e-06, "loss": 0.0735, "step": 6776 }, { "epoch": 2.1, "learning_rate": 7.292258451690338e-06, "loss": 0.0965, "step": 6777 }, { "epoch": 2.1, "learning_rate": 7.291858371674335e-06, "loss": 0.0389, "step": 6778 }, { "epoch": 2.1, "learning_rate": 7.291458291658332e-06, "loss": 0.0772, "step": 6779 }, { "epoch": 2.1, "learning_rate": 7.291058211642329e-06, "loss": 0.0632, "step": 6780 }, { "epoch": 2.1, "learning_rate": 7.290658131626326e-06, "loss": 0.0909, "step": 6781 }, { "epoch": 2.1, "learning_rate": 7.290258051610323e-06, "loss": 0.1025, "step": 6782 }, { "epoch": 2.11, "learning_rate": 7.289857971594319e-06, "loss": 0.1036, "step": 6783 }, { "epoch": 2.11, "learning_rate": 7.289457891578317e-06, "loss": 0.2043, "step": 6784 }, { "epoch": 2.11, "learning_rate": 7.289057811562314e-06, "loss": 0.1017, "step": 6785 }, { "epoch": 2.11, "learning_rate": 7.28865773154631e-06, "loss": 0.0759, "step": 6786 }, { "epoch": 2.11, "learning_rate": 7.288257651530307e-06, "loss": 0.0722, "step": 6787 }, { "epoch": 2.11, "learning_rate": 7.287857571514303e-06, "loss": 0.0976, "step": 6788 }, { "epoch": 2.11, "learning_rate": 7.2874574914982995e-06, "loss": 0.0892, "step": 6789 }, { "epoch": 2.11, "learning_rate": 7.287057411482297e-06, "loss": 0.0847, "step": 6790 }, { "epoch": 2.11, "learning_rate": 7.286657331466294e-06, "loss": 0.1236, "step": 6791 }, { "epoch": 2.11, "learning_rate": 7.28625725145029e-06, "loss": 0.0369, "step": 6792 }, { "epoch": 2.11, "learning_rate": 7.285857171434287e-06, "loss": 0.1173, "step": 6793 }, { "epoch": 2.11, "learning_rate": 7.285457091418285e-06, "loss": 0.0952, "step": 6794 }, { "epoch": 2.11, "learning_rate": 7.285057011402281e-06, "loss": 0.1982, "step": 6795 }, { "epoch": 2.11, "learning_rate": 7.284656931386278e-06, "loss": 0.0686, "step": 6796 }, { "epoch": 2.11, "learning_rate": 7.284256851370275e-06, "loss": 0.0914, "step": 6797 }, { "epoch": 2.11, "learning_rate": 7.283856771354272e-06, "loss": 0.0318, "step": 6798 }, { "epoch": 2.11, "learning_rate": 7.283456691338269e-06, "loss": 0.1287, "step": 6799 }, { "epoch": 2.11, "learning_rate": 7.283056611322266e-06, "loss": 0.0493, "step": 6800 }, { "epoch": 2.11, "learning_rate": 7.282656531306261e-06, "loss": 0.0798, "step": 6801 }, { "epoch": 2.11, "learning_rate": 7.282256451290258e-06, "loss": 0.0559, "step": 6802 }, { "epoch": 2.11, "learning_rate": 7.281856371274255e-06, "loss": 0.0642, "step": 6803 }, { "epoch": 2.11, "learning_rate": 7.281456291258252e-06, "loss": 0.0725, "step": 6804 }, { "epoch": 2.11, "learning_rate": 7.281056211242249e-06, "loss": 0.0737, "step": 6805 }, { "epoch": 2.11, "learning_rate": 7.280656131226246e-06, "loss": 0.1025, "step": 6806 }, { "epoch": 2.11, "learning_rate": 7.280256051210242e-06, "loss": 0.1858, "step": 6807 }, { "epoch": 2.11, "learning_rate": 7.27985597119424e-06, "loss": 0.0627, "step": 6808 }, { "epoch": 2.11, "learning_rate": 7.279455891178237e-06, "loss": 0.213, "step": 6809 }, { "epoch": 2.11, "learning_rate": 7.279055811162233e-06, "loss": 0.075, "step": 6810 }, { "epoch": 2.11, "learning_rate": 7.27865573114623e-06, "loss": 0.0722, "step": 6811 }, { "epoch": 2.11, "learning_rate": 7.2782556511302275e-06, "loss": 0.0201, "step": 6812 }, { "epoch": 2.11, "learning_rate": 7.2778555711142225e-06, "loss": 0.1196, "step": 6813 }, { "epoch": 2.11, "learning_rate": 7.27745549109822e-06, "loss": 0.1898, "step": 6814 }, { "epoch": 2.12, "learning_rate": 7.277055411082217e-06, "loss": 0.065, "step": 6815 }, { "epoch": 2.12, "learning_rate": 7.276655331066213e-06, "loss": 0.1502, "step": 6816 }, { "epoch": 2.12, "learning_rate": 7.27625525105021e-06, "loss": 0.0748, "step": 6817 }, { "epoch": 2.12, "learning_rate": 7.275855171034208e-06, "loss": 0.0491, "step": 6818 }, { "epoch": 2.12, "learning_rate": 7.275455091018204e-06, "loss": 0.2087, "step": 6819 }, { "epoch": 2.12, "learning_rate": 7.275055011002201e-06, "loss": 0.127, "step": 6820 }, { "epoch": 2.12, "learning_rate": 7.274654930986198e-06, "loss": 0.0468, "step": 6821 }, { "epoch": 2.12, "learning_rate": 7.274254850970195e-06, "loss": 0.0466, "step": 6822 }, { "epoch": 2.12, "learning_rate": 7.273854770954192e-06, "loss": 0.0954, "step": 6823 }, { "epoch": 2.12, "learning_rate": 7.273454690938189e-06, "loss": 0.1439, "step": 6824 }, { "epoch": 2.12, "learning_rate": 7.273054610922185e-06, "loss": 0.082, "step": 6825 }, { "epoch": 2.12, "learning_rate": 7.272654530906181e-06, "loss": 0.0943, "step": 6826 }, { "epoch": 2.12, "learning_rate": 7.272254450890179e-06, "loss": 0.0662, "step": 6827 }, { "epoch": 2.12, "learning_rate": 7.271854370874175e-06, "loss": 0.0248, "step": 6828 }, { "epoch": 2.12, "learning_rate": 7.271454290858172e-06, "loss": 0.1163, "step": 6829 }, { "epoch": 2.12, "learning_rate": 7.271054210842169e-06, "loss": 0.0841, "step": 6830 }, { "epoch": 2.12, "learning_rate": 7.270654130826165e-06, "loss": 0.1101, "step": 6831 }, { "epoch": 2.12, "learning_rate": 7.270254050810163e-06, "loss": 0.1006, "step": 6832 }, { "epoch": 2.12, "learning_rate": 7.26985397079416e-06, "loss": 0.0573, "step": 6833 }, { "epoch": 2.12, "learning_rate": 7.269453890778156e-06, "loss": 0.0693, "step": 6834 }, { "epoch": 2.12, "learning_rate": 7.269053810762153e-06, "loss": 0.2168, "step": 6835 }, { "epoch": 2.12, "learning_rate": 7.2686537307461505e-06, "loss": 0.0945, "step": 6836 }, { "epoch": 2.12, "learning_rate": 7.268253650730147e-06, "loss": 0.0414, "step": 6837 }, { "epoch": 2.12, "learning_rate": 7.267853570714143e-06, "loss": 0.0964, "step": 6838 }, { "epoch": 2.12, "learning_rate": 7.26745349069814e-06, "loss": 0.0966, "step": 6839 }, { "epoch": 2.12, "learning_rate": 7.267053410682136e-06, "loss": 0.0925, "step": 6840 }, { "epoch": 2.12, "learning_rate": 7.266653330666134e-06, "loss": 0.0813, "step": 6841 }, { "epoch": 2.12, "learning_rate": 7.266253250650131e-06, "loss": 0.0868, "step": 6842 }, { "epoch": 2.12, "learning_rate": 7.265853170634127e-06, "loss": 0.0622, "step": 6843 }, { "epoch": 2.12, "learning_rate": 7.265453090618124e-06, "loss": 0.1233, "step": 6844 }, { "epoch": 2.12, "learning_rate": 7.265053010602121e-06, "loss": 0.0671, "step": 6845 }, { "epoch": 2.12, "learning_rate": 7.264652930586118e-06, "loss": 0.0956, "step": 6846 }, { "epoch": 2.13, "learning_rate": 7.264252850570115e-06, "loss": 0.0841, "step": 6847 }, { "epoch": 2.13, "learning_rate": 7.263852770554112e-06, "loss": 0.0912, "step": 6848 }, { "epoch": 2.13, "learning_rate": 7.263452690538108e-06, "loss": 0.0576, "step": 6849 }, { "epoch": 2.13, "learning_rate": 7.263052610522106e-06, "loss": 0.0577, "step": 6850 }, { "epoch": 2.13, "learning_rate": 7.262652530506102e-06, "loss": 0.0793, "step": 6851 }, { "epoch": 2.13, "learning_rate": 7.262252450490098e-06, "loss": 0.0719, "step": 6852 }, { "epoch": 2.13, "learning_rate": 7.261852370474095e-06, "loss": 0.0614, "step": 6853 }, { "epoch": 2.13, "learning_rate": 7.261452290458092e-06, "loss": 0.1604, "step": 6854 }, { "epoch": 2.13, "learning_rate": 7.261052210442089e-06, "loss": 0.074, "step": 6855 }, { "epoch": 2.13, "learning_rate": 7.260652130426086e-06, "loss": 0.0348, "step": 6856 }, { "epoch": 2.13, "learning_rate": 7.260252050410083e-06, "loss": 0.0406, "step": 6857 }, { "epoch": 2.13, "learning_rate": 7.259851970394079e-06, "loss": 0.2473, "step": 6858 }, { "epoch": 2.13, "learning_rate": 7.259451890378076e-06, "loss": 0.0287, "step": 6859 }, { "epoch": 2.13, "learning_rate": 7.2590518103620735e-06, "loss": 0.1063, "step": 6860 }, { "epoch": 2.13, "learning_rate": 7.25865173034607e-06, "loss": 0.1159, "step": 6861 }, { "epoch": 2.13, "learning_rate": 7.258251650330067e-06, "loss": 0.1017, "step": 6862 }, { "epoch": 2.13, "learning_rate": 7.257851570314064e-06, "loss": 0.1067, "step": 6863 }, { "epoch": 2.13, "learning_rate": 7.2574514902980594e-06, "loss": 0.0927, "step": 6864 }, { "epoch": 2.13, "learning_rate": 7.257051410282057e-06, "loss": 0.1042, "step": 6865 }, { "epoch": 2.13, "learning_rate": 7.256651330266054e-06, "loss": 0.1709, "step": 6866 }, { "epoch": 2.13, "learning_rate": 7.25625125025005e-06, "loss": 0.0679, "step": 6867 }, { "epoch": 2.13, "learning_rate": 7.255851170234047e-06, "loss": 0.0442, "step": 6868 }, { "epoch": 2.13, "learning_rate": 7.2554510902180446e-06, "loss": 0.083, "step": 6869 }, { "epoch": 2.13, "learning_rate": 7.255051010202041e-06, "loss": 0.0956, "step": 6870 }, { "epoch": 2.13, "learning_rate": 7.254650930186038e-06, "loss": 0.0976, "step": 6871 }, { "epoch": 2.13, "learning_rate": 7.254250850170035e-06, "loss": 0.1155, "step": 6872 }, { "epoch": 2.13, "learning_rate": 7.253850770154031e-06, "loss": 0.0644, "step": 6873 }, { "epoch": 2.13, "learning_rate": 7.253450690138029e-06, "loss": 0.0623, "step": 6874 }, { "epoch": 2.13, "learning_rate": 7.2530506101220255e-06, "loss": 0.1461, "step": 6875 }, { "epoch": 2.13, "learning_rate": 7.252650530106021e-06, "loss": 0.1373, "step": 6876 }, { "epoch": 2.13, "learning_rate": 7.252250450090018e-06, "loss": 0.0406, "step": 6877 }, { "epoch": 2.13, "learning_rate": 7.251850370074015e-06, "loss": 0.0876, "step": 6878 }, { "epoch": 2.14, "learning_rate": 7.251450290058012e-06, "loss": 0.0741, "step": 6879 }, { "epoch": 2.14, "learning_rate": 7.251050210042009e-06, "loss": 0.1394, "step": 6880 }, { "epoch": 2.14, "learning_rate": 7.250650130026006e-06, "loss": 0.1033, "step": 6881 }, { "epoch": 2.14, "learning_rate": 7.250250050010002e-06, "loss": 0.117, "step": 6882 }, { "epoch": 2.14, "learning_rate": 7.249849969994e-06, "loss": 0.0611, "step": 6883 }, { "epoch": 2.14, "learning_rate": 7.2494498899779965e-06, "loss": 0.099, "step": 6884 }, { "epoch": 2.14, "learning_rate": 7.249049809961993e-06, "loss": 0.1253, "step": 6885 }, { "epoch": 2.14, "learning_rate": 7.24864972994599e-06, "loss": 0.0647, "step": 6886 }, { "epoch": 2.14, "learning_rate": 7.2482496499299875e-06, "loss": 0.0507, "step": 6887 }, { "epoch": 2.14, "learning_rate": 7.247849569913984e-06, "loss": 0.074, "step": 6888 }, { "epoch": 2.14, "learning_rate": 7.24744948989798e-06, "loss": 0.1038, "step": 6889 }, { "epoch": 2.14, "learning_rate": 7.247049409881977e-06, "loss": 0.0246, "step": 6890 }, { "epoch": 2.14, "learning_rate": 7.246649329865973e-06, "loss": 0.2207, "step": 6891 }, { "epoch": 2.14, "learning_rate": 7.24624924984997e-06, "loss": 0.1708, "step": 6892 }, { "epoch": 2.14, "learning_rate": 7.2458491698339676e-06, "loss": 0.1178, "step": 6893 }, { "epoch": 2.14, "learning_rate": 7.245449089817964e-06, "loss": 0.1426, "step": 6894 }, { "epoch": 2.14, "learning_rate": 7.245049009801961e-06, "loss": 0.059, "step": 6895 }, { "epoch": 2.14, "learning_rate": 7.244648929785958e-06, "loss": 0.1338, "step": 6896 }, { "epoch": 2.14, "learning_rate": 7.244248849769955e-06, "loss": 0.1147, "step": 6897 }, { "epoch": 2.14, "learning_rate": 7.243848769753952e-06, "loss": 0.1046, "step": 6898 }, { "epoch": 2.14, "learning_rate": 7.2434486897379485e-06, "loss": 0.3268, "step": 6899 }, { "epoch": 2.14, "learning_rate": 7.243048609721945e-06, "loss": 0.0702, "step": 6900 }, { "epoch": 2.14, "learning_rate": 7.242648529705941e-06, "loss": 0.068, "step": 6901 }, { "epoch": 2.14, "learning_rate": 7.242248449689938e-06, "loss": 0.0606, "step": 6902 }, { "epoch": 2.14, "learning_rate": 7.241848369673935e-06, "loss": 0.0511, "step": 6903 }, { "epoch": 2.14, "learning_rate": 7.241448289657932e-06, "loss": 0.0791, "step": 6904 }, { "epoch": 2.14, "learning_rate": 7.241048209641929e-06, "loss": 0.0614, "step": 6905 }, { "epoch": 2.14, "learning_rate": 7.240648129625925e-06, "loss": 0.0523, "step": 6906 }, { "epoch": 2.14, "learning_rate": 7.240248049609923e-06, "loss": 0.1309, "step": 6907 }, { "epoch": 2.14, "learning_rate": 7.2398479695939196e-06, "loss": 0.0897, "step": 6908 }, { "epoch": 2.14, "learning_rate": 7.239447889577916e-06, "loss": 0.1068, "step": 6909 }, { "epoch": 2.14, "learning_rate": 7.239047809561913e-06, "loss": 0.062, "step": 6910 }, { "epoch": 2.15, "learning_rate": 7.2386477295459105e-06, "loss": 0.0664, "step": 6911 }, { "epoch": 2.15, "learning_rate": 7.238247649529907e-06, "loss": 0.121, "step": 6912 }, { "epoch": 2.15, "learning_rate": 7.237847569513904e-06, "loss": 0.0914, "step": 6913 }, { "epoch": 2.15, "learning_rate": 7.2374474894979e-06, "loss": 0.1257, "step": 6914 }, { "epoch": 2.15, "learning_rate": 7.237047409481896e-06, "loss": 0.0395, "step": 6915 }, { "epoch": 2.15, "learning_rate": 7.236647329465893e-06, "loss": 0.1411, "step": 6916 }, { "epoch": 2.15, "learning_rate": 7.236247249449891e-06, "loss": 0.0572, "step": 6917 }, { "epoch": 2.15, "learning_rate": 7.235847169433887e-06, "loss": 0.0785, "step": 6918 }, { "epoch": 2.15, "learning_rate": 7.235447089417884e-06, "loss": 0.05, "step": 6919 }, { "epoch": 2.15, "learning_rate": 7.235047009401881e-06, "loss": 0.0657, "step": 6920 }, { "epoch": 2.15, "learning_rate": 7.234646929385878e-06, "loss": 0.1742, "step": 6921 }, { "epoch": 2.15, "learning_rate": 7.234246849369875e-06, "loss": 0.0734, "step": 6922 }, { "epoch": 2.15, "learning_rate": 7.2338467693538715e-06, "loss": 0.0482, "step": 6923 }, { "epoch": 2.15, "learning_rate": 7.233446689337868e-06, "loss": 0.1113, "step": 6924 }, { "epoch": 2.15, "learning_rate": 7.233046609321866e-06, "loss": 0.0647, "step": 6925 }, { "epoch": 2.15, "learning_rate": 7.2326465293058625e-06, "loss": 0.1749, "step": 6926 }, { "epoch": 2.15, "learning_rate": 7.232246449289858e-06, "loss": 0.041, "step": 6927 }, { "epoch": 2.15, "learning_rate": 7.231846369273855e-06, "loss": 0.0537, "step": 6928 }, { "epoch": 2.15, "learning_rate": 7.231446289257852e-06, "loss": 0.1222, "step": 6929 }, { "epoch": 2.15, "learning_rate": 7.231046209241848e-06, "loss": 0.2086, "step": 6930 }, { "epoch": 2.15, "learning_rate": 7.230646129225846e-06, "loss": 0.1135, "step": 6931 }, { "epoch": 2.15, "learning_rate": 7.2302460492098426e-06, "loss": 0.1253, "step": 6932 }, { "epoch": 2.15, "learning_rate": 7.229845969193839e-06, "loss": 0.1097, "step": 6933 }, { "epoch": 2.15, "learning_rate": 7.229445889177836e-06, "loss": 0.0678, "step": 6934 }, { "epoch": 2.15, "learning_rate": 7.2290458091618335e-06, "loss": 0.1218, "step": 6935 }, { "epoch": 2.15, "learning_rate": 7.22864572914583e-06, "loss": 0.0601, "step": 6936 }, { "epoch": 2.15, "learning_rate": 7.228245649129827e-06, "loss": 0.09, "step": 6937 }, { "epoch": 2.15, "learning_rate": 7.2278455691138235e-06, "loss": 0.0644, "step": 6938 }, { "epoch": 2.15, "learning_rate": 7.227445489097819e-06, "loss": 0.061, "step": 6939 }, { "epoch": 2.15, "learning_rate": 7.227045409081817e-06, "loss": 0.121, "step": 6940 }, { "epoch": 2.15, "learning_rate": 7.226645329065814e-06, "loss": 0.1149, "step": 6941 }, { "epoch": 2.15, "learning_rate": 7.22624524904981e-06, "loss": 0.0838, "step": 6942 }, { "epoch": 2.15, "learning_rate": 7.225845169033807e-06, "loss": 0.0893, "step": 6943 }, { "epoch": 2.16, "learning_rate": 7.225445089017804e-06, "loss": 0.0909, "step": 6944 }, { "epoch": 2.16, "learning_rate": 7.225045009001801e-06, "loss": 0.0934, "step": 6945 }, { "epoch": 2.16, "learning_rate": 7.224644928985798e-06, "loss": 0.2084, "step": 6946 }, { "epoch": 2.16, "learning_rate": 7.2242448489697946e-06, "loss": 0.0775, "step": 6947 }, { "epoch": 2.16, "learning_rate": 7.223844768953791e-06, "loss": 0.0609, "step": 6948 }, { "epoch": 2.16, "learning_rate": 7.223444688937789e-06, "loss": 0.0626, "step": 6949 }, { "epoch": 2.16, "learning_rate": 7.2230446089217855e-06, "loss": 0.1423, "step": 6950 }, { "epoch": 2.16, "learning_rate": 7.222644528905782e-06, "loss": 0.0589, "step": 6951 }, { "epoch": 2.16, "learning_rate": 7.222244448889778e-06, "loss": 0.2454, "step": 6952 }, { "epoch": 2.16, "learning_rate": 7.221844368873775e-06, "loss": 0.0962, "step": 6953 }, { "epoch": 2.16, "learning_rate": 7.221444288857772e-06, "loss": 0.0634, "step": 6954 }, { "epoch": 2.16, "learning_rate": 7.221044208841769e-06, "loss": 0.082, "step": 6955 }, { "epoch": 2.16, "learning_rate": 7.220644128825766e-06, "loss": 0.0906, "step": 6956 }, { "epoch": 2.16, "learning_rate": 7.220244048809762e-06, "loss": 0.1205, "step": 6957 }, { "epoch": 2.16, "learning_rate": 7.219843968793759e-06, "loss": 0.1236, "step": 6958 }, { "epoch": 2.16, "learning_rate": 7.2194438887777565e-06, "loss": 0.0882, "step": 6959 }, { "epoch": 2.16, "learning_rate": 7.219043808761753e-06, "loss": 0.1091, "step": 6960 }, { "epoch": 2.16, "learning_rate": 7.21864372874575e-06, "loss": 0.077, "step": 6961 }, { "epoch": 2.16, "learning_rate": 7.2182436487297465e-06, "loss": 0.1176, "step": 6962 }, { "epoch": 2.16, "learning_rate": 7.217843568713744e-06, "loss": 0.1703, "step": 6963 }, { "epoch": 2.16, "learning_rate": 7.21744348869774e-06, "loss": 0.1598, "step": 6964 }, { "epoch": 2.16, "learning_rate": 7.217043408681737e-06, "loss": 0.0571, "step": 6965 }, { "epoch": 2.16, "learning_rate": 7.216643328665733e-06, "loss": 0.0546, "step": 6966 }, { "epoch": 2.16, "learning_rate": 7.21624324864973e-06, "loss": 0.0582, "step": 6967 }, { "epoch": 2.16, "learning_rate": 7.2158431686337275e-06, "loss": 0.0949, "step": 6968 }, { "epoch": 2.16, "learning_rate": 7.215443088617724e-06, "loss": 0.107, "step": 6969 }, { "epoch": 2.16, "learning_rate": 7.215043008601721e-06, "loss": 0.1071, "step": 6970 }, { "epoch": 2.16, "learning_rate": 7.2146429285857176e-06, "loss": 0.0355, "step": 6971 }, { "epoch": 2.16, "learning_rate": 7.214242848569715e-06, "loss": 0.0668, "step": 6972 }, { "epoch": 2.16, "learning_rate": 7.213842768553712e-06, "loss": 0.0817, "step": 6973 }, { "epoch": 2.16, "learning_rate": 7.2134426885377085e-06, "loss": 0.0462, "step": 6974 }, { "epoch": 2.16, "learning_rate": 7.213042608521705e-06, "loss": 0.0856, "step": 6975 }, { "epoch": 2.17, "learning_rate": 7.212642528505702e-06, "loss": 0.0772, "step": 6976 }, { "epoch": 2.17, "learning_rate": 7.212242448489698e-06, "loss": 0.0756, "step": 6977 }, { "epoch": 2.17, "learning_rate": 7.211842368473695e-06, "loss": 0.1512, "step": 6978 }, { "epoch": 2.17, "learning_rate": 7.211442288457692e-06, "loss": 0.0663, "step": 6979 }, { "epoch": 2.17, "learning_rate": 7.211042208441689e-06, "loss": 0.0998, "step": 6980 }, { "epoch": 2.17, "learning_rate": 7.210642128425685e-06, "loss": 0.1251, "step": 6981 }, { "epoch": 2.17, "learning_rate": 7.210242048409683e-06, "loss": 0.1192, "step": 6982 }, { "epoch": 2.17, "learning_rate": 7.2098419683936795e-06, "loss": 0.1013, "step": 6983 }, { "epoch": 2.17, "learning_rate": 7.209441888377676e-06, "loss": 0.0637, "step": 6984 }, { "epoch": 2.17, "learning_rate": 7.209041808361673e-06, "loss": 0.1127, "step": 6985 }, { "epoch": 2.17, "learning_rate": 7.20864172834567e-06, "loss": 0.0606, "step": 6986 }, { "epoch": 2.17, "learning_rate": 7.208241648329667e-06, "loss": 0.0958, "step": 6987 }, { "epoch": 2.17, "learning_rate": 7.207841568313664e-06, "loss": 0.0846, "step": 6988 }, { "epoch": 2.17, "learning_rate": 7.2074414882976605e-06, "loss": 0.0541, "step": 6989 }, { "epoch": 2.17, "learning_rate": 7.207041408281656e-06, "loss": 0.0576, "step": 6990 }, { "epoch": 2.17, "learning_rate": 7.206641328265653e-06, "loss": 0.1157, "step": 6991 }, { "epoch": 2.17, "learning_rate": 7.2062412482496505e-06, "loss": 0.0664, "step": 6992 }, { "epoch": 2.17, "learning_rate": 7.205841168233647e-06, "loss": 0.1094, "step": 6993 }, { "epoch": 2.17, "learning_rate": 7.205441088217644e-06, "loss": 0.0336, "step": 6994 }, { "epoch": 2.17, "learning_rate": 7.205041008201641e-06, "loss": 0.0726, "step": 6995 }, { "epoch": 2.17, "learning_rate": 7.204640928185638e-06, "loss": 0.0716, "step": 6996 }, { "epoch": 2.17, "learning_rate": 7.204240848169635e-06, "loss": 0.082, "step": 6997 }, { "epoch": 2.17, "learning_rate": 7.2038407681536315e-06, "loss": 0.1087, "step": 6998 }, { "epoch": 2.17, "learning_rate": 7.203440688137628e-06, "loss": 0.0855, "step": 6999 }, { "epoch": 2.17, "learning_rate": 7.203040608121626e-06, "loss": 0.0678, "step": 7000 }, { "epoch": 2.17, "learning_rate": 7.202640528105622e-06, "loss": 0.108, "step": 7001 }, { "epoch": 2.17, "learning_rate": 7.202240448089618e-06, "loss": 0.1199, "step": 7002 }, { "epoch": 2.17, "learning_rate": 7.201840368073615e-06, "loss": 0.0439, "step": 7003 }, { "epoch": 2.17, "learning_rate": 7.201440288057612e-06, "loss": 0.1437, "step": 7004 }, { "epoch": 2.17, "learning_rate": 7.201040208041608e-06, "loss": 0.0749, "step": 7005 }, { "epoch": 2.17, "learning_rate": 7.200640128025606e-06, "loss": 0.0709, "step": 7006 }, { "epoch": 2.17, "learning_rate": 7.2002400480096025e-06, "loss": 0.1414, "step": 7007 }, { "epoch": 2.18, "learning_rate": 7.199839967993599e-06, "loss": 0.1328, "step": 7008 }, { "epoch": 2.18, "learning_rate": 7.199439887977596e-06, "loss": 0.0413, "step": 7009 }, { "epoch": 2.18, "learning_rate": 7.199039807961593e-06, "loss": 0.0817, "step": 7010 }, { "epoch": 2.18, "learning_rate": 7.19863972794559e-06, "loss": 0.0672, "step": 7011 }, { "epoch": 2.18, "learning_rate": 7.198239647929587e-06, "loss": 0.0671, "step": 7012 }, { "epoch": 2.18, "learning_rate": 7.1978395679135835e-06, "loss": 0.0841, "step": 7013 }, { "epoch": 2.18, "learning_rate": 7.197439487897581e-06, "loss": 0.1742, "step": 7014 }, { "epoch": 2.18, "learning_rate": 7.197039407881576e-06, "loss": 0.0518, "step": 7015 }, { "epoch": 2.18, "learning_rate": 7.1966393278655735e-06, "loss": 0.1422, "step": 7016 }, { "epoch": 2.18, "learning_rate": 7.19623924784957e-06, "loss": 0.0811, "step": 7017 }, { "epoch": 2.18, "learning_rate": 7.195839167833567e-06, "loss": 0.111, "step": 7018 }, { "epoch": 2.18, "learning_rate": 7.195439087817564e-06, "loss": 0.0721, "step": 7019 }, { "epoch": 2.18, "learning_rate": 7.195039007801561e-06, "loss": 0.0633, "step": 7020 }, { "epoch": 2.18, "learning_rate": 7.194638927785558e-06, "loss": 0.0894, "step": 7021 }, { "epoch": 2.18, "learning_rate": 7.1942388477695545e-06, "loss": 0.1052, "step": 7022 }, { "epoch": 2.18, "learning_rate": 7.193838767753551e-06, "loss": 0.0684, "step": 7023 }, { "epoch": 2.18, "learning_rate": 7.193438687737549e-06, "loss": 0.1056, "step": 7024 }, { "epoch": 2.18, "learning_rate": 7.193038607721545e-06, "loss": 0.1007, "step": 7025 }, { "epoch": 2.18, "learning_rate": 7.192638527705542e-06, "loss": 0.1169, "step": 7026 }, { "epoch": 2.18, "learning_rate": 7.192238447689539e-06, "loss": 0.0812, "step": 7027 }, { "epoch": 2.18, "learning_rate": 7.191838367673535e-06, "loss": 0.0682, "step": 7028 }, { "epoch": 2.18, "learning_rate": 7.191438287657531e-06, "loss": 0.1223, "step": 7029 }, { "epoch": 2.18, "learning_rate": 7.191038207641529e-06, "loss": 0.1002, "step": 7030 }, { "epoch": 2.18, "learning_rate": 7.1906381276255255e-06, "loss": 0.0354, "step": 7031 }, { "epoch": 2.18, "learning_rate": 7.190238047609522e-06, "loss": 0.0675, "step": 7032 }, { "epoch": 2.18, "learning_rate": 7.189837967593519e-06, "loss": 0.1105, "step": 7033 }, { "epoch": 2.18, "learning_rate": 7.1894378875775164e-06, "loss": 0.0751, "step": 7034 }, { "epoch": 2.18, "learning_rate": 7.189037807561513e-06, "loss": 0.0442, "step": 7035 }, { "epoch": 2.18, "learning_rate": 7.18863772754551e-06, "loss": 0.0781, "step": 7036 }, { "epoch": 2.18, "learning_rate": 7.1882376475295065e-06, "loss": 0.0703, "step": 7037 }, { "epoch": 2.18, "learning_rate": 7.187837567513504e-06, "loss": 0.1024, "step": 7038 }, { "epoch": 2.18, "learning_rate": 7.187437487497501e-06, "loss": 0.0874, "step": 7039 }, { "epoch": 2.19, "learning_rate": 7.1870374074814965e-06, "loss": 0.0578, "step": 7040 }, { "epoch": 2.19, "learning_rate": 7.186637327465493e-06, "loss": 0.0945, "step": 7041 }, { "epoch": 2.19, "learning_rate": 7.18623724744949e-06, "loss": 0.0713, "step": 7042 }, { "epoch": 2.19, "learning_rate": 7.185837167433487e-06, "loss": 0.1103, "step": 7043 }, { "epoch": 2.19, "learning_rate": 7.185437087417484e-06, "loss": 0.1058, "step": 7044 }, { "epoch": 2.19, "learning_rate": 7.185037007401481e-06, "loss": 0.1172, "step": 7045 }, { "epoch": 2.19, "learning_rate": 7.1846369273854775e-06, "loss": 0.0445, "step": 7046 }, { "epoch": 2.19, "learning_rate": 7.184236847369474e-06, "loss": 0.1306, "step": 7047 }, { "epoch": 2.19, "learning_rate": 7.183836767353472e-06, "loss": 0.0367, "step": 7048 }, { "epoch": 2.19, "learning_rate": 7.183436687337468e-06, "loss": 0.11, "step": 7049 }, { "epoch": 2.19, "learning_rate": 7.183036607321465e-06, "loss": 0.0955, "step": 7050 }, { "epoch": 2.19, "learning_rate": 7.182636527305462e-06, "loss": 0.1228, "step": 7051 }, { "epoch": 2.19, "learning_rate": 7.182236447289459e-06, "loss": 0.0568, "step": 7052 }, { "epoch": 2.19, "learning_rate": 7.181836367273455e-06, "loss": 0.115, "step": 7053 }, { "epoch": 2.19, "learning_rate": 7.181436287257452e-06, "loss": 0.1153, "step": 7054 }, { "epoch": 2.19, "learning_rate": 7.1810362072414485e-06, "loss": 0.1292, "step": 7055 }, { "epoch": 2.19, "learning_rate": 7.180636127225445e-06, "loss": 0.043, "step": 7056 }, { "epoch": 2.19, "learning_rate": 7.180236047209443e-06, "loss": 0.2088, "step": 7057 }, { "epoch": 2.19, "learning_rate": 7.1798359671934394e-06, "loss": 0.0661, "step": 7058 }, { "epoch": 2.19, "learning_rate": 7.179435887177436e-06, "loss": 0.0495, "step": 7059 }, { "epoch": 2.19, "learning_rate": 7.179035807161433e-06, "loss": 0.0482, "step": 7060 }, { "epoch": 2.19, "learning_rate": 7.1786357271454295e-06, "loss": 0.1441, "step": 7061 }, { "epoch": 2.19, "learning_rate": 7.178235647129427e-06, "loss": 0.0405, "step": 7062 }, { "epoch": 2.19, "learning_rate": 7.177835567113424e-06, "loss": 0.1183, "step": 7063 }, { "epoch": 2.19, "learning_rate": 7.17743548709742e-06, "loss": 0.0931, "step": 7064 }, { "epoch": 2.19, "learning_rate": 7.177035407081416e-06, "loss": 0.1536, "step": 7065 }, { "epoch": 2.19, "learning_rate": 7.176635327065413e-06, "loss": 0.0853, "step": 7066 }, { "epoch": 2.19, "learning_rate": 7.1762352470494105e-06, "loss": 0.0474, "step": 7067 }, { "epoch": 2.19, "learning_rate": 7.175835167033407e-06, "loss": 0.1793, "step": 7068 }, { "epoch": 2.19, "learning_rate": 7.175435087017404e-06, "loss": 0.0354, "step": 7069 }, { "epoch": 2.19, "learning_rate": 7.1750350070014005e-06, "loss": 0.0917, "step": 7070 }, { "epoch": 2.19, "learning_rate": 7.174634926985398e-06, "loss": 0.0849, "step": 7071 }, { "epoch": 2.19, "learning_rate": 7.174234846969395e-06, "loss": 0.0459, "step": 7072 }, { "epoch": 2.2, "learning_rate": 7.1738347669533914e-06, "loss": 0.1314, "step": 7073 }, { "epoch": 2.2, "learning_rate": 7.173434686937388e-06, "loss": 0.0727, "step": 7074 }, { "epoch": 2.2, "learning_rate": 7.173034606921385e-06, "loss": 0.083, "step": 7075 }, { "epoch": 2.2, "learning_rate": 7.172634526905382e-06, "loss": 0.0994, "step": 7076 }, { "epoch": 2.2, "learning_rate": 7.172234446889379e-06, "loss": 0.0321, "step": 7077 }, { "epoch": 2.2, "learning_rate": 7.171834366873375e-06, "loss": 0.0685, "step": 7078 }, { "epoch": 2.2, "learning_rate": 7.1714342868573715e-06, "loss": 0.1, "step": 7079 }, { "epoch": 2.2, "learning_rate": 7.171034206841368e-06, "loss": 0.0532, "step": 7080 }, { "epoch": 2.2, "learning_rate": 7.170634126825366e-06, "loss": 0.142, "step": 7081 }, { "epoch": 2.2, "learning_rate": 7.1702340468093625e-06, "loss": 0.0602, "step": 7082 }, { "epoch": 2.2, "learning_rate": 7.169833966793359e-06, "loss": 0.0652, "step": 7083 }, { "epoch": 2.2, "learning_rate": 7.169433886777356e-06, "loss": 0.1078, "step": 7084 }, { "epoch": 2.2, "learning_rate": 7.169033806761353e-06, "loss": 0.0918, "step": 7085 }, { "epoch": 2.2, "learning_rate": 7.16863372674535e-06, "loss": 0.098, "step": 7086 }, { "epoch": 2.2, "learning_rate": 7.168233646729347e-06, "loss": 0.1762, "step": 7087 }, { "epoch": 2.2, "learning_rate": 7.167833566713343e-06, "loss": 0.1207, "step": 7088 }, { "epoch": 2.2, "learning_rate": 7.16743348669734e-06, "loss": 0.0379, "step": 7089 }, { "epoch": 2.2, "learning_rate": 7.167033406681338e-06, "loss": 0.0448, "step": 7090 }, { "epoch": 2.2, "learning_rate": 7.1666333266653335e-06, "loss": 0.0522, "step": 7091 }, { "epoch": 2.2, "learning_rate": 7.16623324664933e-06, "loss": 0.1119, "step": 7092 }, { "epoch": 2.2, "learning_rate": 7.165833166633327e-06, "loss": 0.215, "step": 7093 }, { "epoch": 2.2, "learning_rate": 7.1654330866173235e-06, "loss": 0.1048, "step": 7094 }, { "epoch": 2.2, "learning_rate": 7.165033006601321e-06, "loss": 0.0885, "step": 7095 }, { "epoch": 2.2, "learning_rate": 7.164632926585318e-06, "loss": 0.0752, "step": 7096 }, { "epoch": 2.2, "learning_rate": 7.1642328465693144e-06, "loss": 0.0618, "step": 7097 }, { "epoch": 2.2, "learning_rate": 7.163832766553311e-06, "loss": 0.1062, "step": 7098 }, { "epoch": 2.2, "learning_rate": 7.163432686537309e-06, "loss": 0.0802, "step": 7099 }, { "epoch": 2.2, "learning_rate": 7.163032606521305e-06, "loss": 0.032, "step": 7100 }, { "epoch": 2.2, "learning_rate": 7.162632526505302e-06, "loss": 0.0681, "step": 7101 }, { "epoch": 2.2, "learning_rate": 7.162232446489299e-06, "loss": 0.1074, "step": 7102 }, { "epoch": 2.2, "learning_rate": 7.1618323664732946e-06, "loss": 0.1055, "step": 7103 }, { "epoch": 2.2, "learning_rate": 7.161432286457291e-06, "loss": 0.0947, "step": 7104 }, { "epoch": 2.21, "learning_rate": 7.161032206441289e-06, "loss": 0.0815, "step": 7105 }, { "epoch": 2.21, "learning_rate": 7.1606321264252855e-06, "loss": 0.1841, "step": 7106 }, { "epoch": 2.21, "learning_rate": 7.160232046409282e-06, "loss": 0.0633, "step": 7107 }, { "epoch": 2.21, "learning_rate": 7.159831966393279e-06, "loss": 0.0842, "step": 7108 }, { "epoch": 2.21, "learning_rate": 7.159431886377276e-06, "loss": 0.0594, "step": 7109 }, { "epoch": 2.21, "learning_rate": 7.159031806361273e-06, "loss": 0.0915, "step": 7110 }, { "epoch": 2.21, "learning_rate": 7.15863172634527e-06, "loss": 0.0395, "step": 7111 }, { "epoch": 2.21, "learning_rate": 7.1582316463292664e-06, "loss": 0.1008, "step": 7112 }, { "epoch": 2.21, "learning_rate": 7.157831566313264e-06, "loss": 0.0769, "step": 7113 }, { "epoch": 2.21, "learning_rate": 7.157431486297261e-06, "loss": 0.0825, "step": 7114 }, { "epoch": 2.21, "learning_rate": 7.157031406281257e-06, "loss": 0.0565, "step": 7115 }, { "epoch": 2.21, "learning_rate": 7.156631326265253e-06, "loss": 0.0888, "step": 7116 }, { "epoch": 2.21, "learning_rate": 7.15623124624925e-06, "loss": 0.0659, "step": 7117 }, { "epoch": 2.21, "learning_rate": 7.1558311662332465e-06, "loss": 0.037, "step": 7118 }, { "epoch": 2.21, "learning_rate": 7.155431086217244e-06, "loss": 0.0941, "step": 7119 }, { "epoch": 2.21, "learning_rate": 7.155031006201241e-06, "loss": 0.0867, "step": 7120 }, { "epoch": 2.21, "learning_rate": 7.1546309261852375e-06, "loss": 0.0886, "step": 7121 }, { "epoch": 2.21, "learning_rate": 7.154230846169234e-06, "loss": 0.0458, "step": 7122 }, { "epoch": 2.21, "learning_rate": 7.153830766153232e-06, "loss": 0.082, "step": 7123 }, { "epoch": 2.21, "learning_rate": 7.153430686137228e-06, "loss": 0.046, "step": 7124 }, { "epoch": 2.21, "learning_rate": 7.153030606121225e-06, "loss": 0.0753, "step": 7125 }, { "epoch": 2.21, "learning_rate": 7.152630526105222e-06, "loss": 0.0922, "step": 7126 }, { "epoch": 2.21, "learning_rate": 7.152230446089219e-06, "loss": 0.052, "step": 7127 }, { "epoch": 2.21, "learning_rate": 7.151830366073215e-06, "loss": 0.1254, "step": 7128 }, { "epoch": 2.21, "learning_rate": 7.151430286057212e-06, "loss": 0.073, "step": 7129 }, { "epoch": 2.21, "learning_rate": 7.1510302060412085e-06, "loss": 0.1464, "step": 7130 }, { "epoch": 2.21, "learning_rate": 7.150630126025205e-06, "loss": 0.044, "step": 7131 }, { "epoch": 2.21, "learning_rate": 7.150230046009202e-06, "loss": 0.0862, "step": 7132 }, { "epoch": 2.21, "learning_rate": 7.149829965993199e-06, "loss": 0.078, "step": 7133 }, { "epoch": 2.21, "learning_rate": 7.149429885977196e-06, "loss": 0.1787, "step": 7134 }, { "epoch": 2.21, "learning_rate": 7.149029805961193e-06, "loss": 0.0381, "step": 7135 }, { "epoch": 2.21, "learning_rate": 7.1486297259451894e-06, "loss": 0.115, "step": 7136 }, { "epoch": 2.22, "learning_rate": 7.148229645929187e-06, "loss": 0.1568, "step": 7137 }, { "epoch": 2.22, "learning_rate": 7.147829565913184e-06, "loss": 0.2099, "step": 7138 }, { "epoch": 2.22, "learning_rate": 7.14742948589718e-06, "loss": 0.0828, "step": 7139 }, { "epoch": 2.22, "learning_rate": 7.147029405881177e-06, "loss": 0.0635, "step": 7140 }, { "epoch": 2.22, "learning_rate": 7.146629325865173e-06, "loss": 0.1133, "step": 7141 }, { "epoch": 2.22, "learning_rate": 7.14622924584917e-06, "loss": 0.0999, "step": 7142 }, { "epoch": 2.22, "learning_rate": 7.145829165833167e-06, "loss": 0.0811, "step": 7143 }, { "epoch": 2.22, "learning_rate": 7.145429085817164e-06, "loss": 0.0872, "step": 7144 }, { "epoch": 2.22, "learning_rate": 7.1450290058011605e-06, "loss": 0.1764, "step": 7145 }, { "epoch": 2.22, "learning_rate": 7.144628925785157e-06, "loss": 0.0763, "step": 7146 }, { "epoch": 2.22, "learning_rate": 7.144228845769155e-06, "loss": 0.0509, "step": 7147 }, { "epoch": 2.22, "learning_rate": 7.143828765753151e-06, "loss": 0.0687, "step": 7148 }, { "epoch": 2.22, "learning_rate": 7.143428685737148e-06, "loss": 0.1126, "step": 7149 }, { "epoch": 2.22, "learning_rate": 7.143028605721145e-06, "loss": 0.1641, "step": 7150 }, { "epoch": 2.22, "learning_rate": 7.142628525705142e-06, "loss": 0.1852, "step": 7151 }, { "epoch": 2.22, "learning_rate": 7.142228445689139e-06, "loss": 0.0406, "step": 7152 }, { "epoch": 2.22, "learning_rate": 7.141828365673136e-06, "loss": 0.07, "step": 7153 }, { "epoch": 2.22, "learning_rate": 7.1414282856571315e-06, "loss": 0.04, "step": 7154 }, { "epoch": 2.22, "learning_rate": 7.141028205641128e-06, "loss": 0.2184, "step": 7155 }, { "epoch": 2.22, "learning_rate": 7.140628125625126e-06, "loss": 0.0471, "step": 7156 }, { "epoch": 2.22, "learning_rate": 7.140228045609122e-06, "loss": 0.0647, "step": 7157 }, { "epoch": 2.22, "learning_rate": 7.139827965593119e-06, "loss": 0.0723, "step": 7158 }, { "epoch": 2.22, "learning_rate": 7.139427885577116e-06, "loss": 0.0737, "step": 7159 }, { "epoch": 2.22, "learning_rate": 7.1390278055611125e-06, "loss": 0.1307, "step": 7160 }, { "epoch": 2.22, "learning_rate": 7.13862772554511e-06, "loss": 0.1074, "step": 7161 }, { "epoch": 2.22, "learning_rate": 7.138227645529107e-06, "loss": 0.0586, "step": 7162 }, { "epoch": 2.22, "learning_rate": 7.137827565513103e-06, "loss": 0.116, "step": 7163 }, { "epoch": 2.22, "learning_rate": 7.1374274854971e-06, "loss": 0.0972, "step": 7164 }, { "epoch": 2.22, "learning_rate": 7.1370274054810976e-06, "loss": 0.0877, "step": 7165 }, { "epoch": 2.22, "learning_rate": 7.136627325465093e-06, "loss": 0.087, "step": 7166 }, { "epoch": 2.22, "learning_rate": 7.13622724544909e-06, "loss": 0.0811, "step": 7167 }, { "epoch": 2.22, "learning_rate": 7.135827165433087e-06, "loss": 0.1993, "step": 7168 }, { "epoch": 2.23, "learning_rate": 7.1354270854170835e-06, "loss": 0.1893, "step": 7169 }, { "epoch": 2.23, "learning_rate": 7.135027005401081e-06, "loss": 0.1207, "step": 7170 }, { "epoch": 2.23, "learning_rate": 7.134626925385078e-06, "loss": 0.0455, "step": 7171 }, { "epoch": 2.23, "learning_rate": 7.134226845369074e-06, "loss": 0.152, "step": 7172 }, { "epoch": 2.23, "learning_rate": 7.133826765353071e-06, "loss": 0.0499, "step": 7173 }, { "epoch": 2.23, "learning_rate": 7.133426685337068e-06, "loss": 0.0617, "step": 7174 }, { "epoch": 2.23, "learning_rate": 7.133026605321065e-06, "loss": 0.1341, "step": 7175 }, { "epoch": 2.23, "learning_rate": 7.132626525305062e-06, "loss": 0.1421, "step": 7176 }, { "epoch": 2.23, "learning_rate": 7.132226445289059e-06, "loss": 0.0826, "step": 7177 }, { "epoch": 2.23, "learning_rate": 7.131826365273055e-06, "loss": 0.0905, "step": 7178 }, { "epoch": 2.23, "learning_rate": 7.131426285257051e-06, "loss": 0.0923, "step": 7179 }, { "epoch": 2.23, "learning_rate": 7.131026205241049e-06, "loss": 0.0891, "step": 7180 }, { "epoch": 2.23, "learning_rate": 7.130626125225045e-06, "loss": 0.1189, "step": 7181 }, { "epoch": 2.23, "learning_rate": 7.130226045209042e-06, "loss": 0.032, "step": 7182 }, { "epoch": 2.23, "learning_rate": 7.129825965193039e-06, "loss": 0.0821, "step": 7183 }, { "epoch": 2.23, "learning_rate": 7.129425885177036e-06, "loss": 0.1444, "step": 7184 }, { "epoch": 2.23, "learning_rate": 7.129025805161033e-06, "loss": 0.1223, "step": 7185 }, { "epoch": 2.23, "learning_rate": 7.12862572514503e-06, "loss": 0.0476, "step": 7186 }, { "epoch": 2.23, "learning_rate": 7.128225645129026e-06, "loss": 0.0634, "step": 7187 }, { "epoch": 2.23, "learning_rate": 7.127825565113024e-06, "loss": 0.0965, "step": 7188 }, { "epoch": 2.23, "learning_rate": 7.127425485097021e-06, "loss": 0.1138, "step": 7189 }, { "epoch": 2.23, "learning_rate": 7.127025405081017e-06, "loss": 0.0765, "step": 7190 }, { "epoch": 2.23, "learning_rate": 7.126625325065013e-06, "loss": 0.1112, "step": 7191 }, { "epoch": 2.23, "learning_rate": 7.12622524504901e-06, "loss": 0.0749, "step": 7192 }, { "epoch": 2.23, "learning_rate": 7.1258251650330065e-06, "loss": 0.0861, "step": 7193 }, { "epoch": 2.23, "learning_rate": 7.125425085017004e-06, "loss": 0.1091, "step": 7194 }, { "epoch": 2.23, "learning_rate": 7.125025005001001e-06, "loss": 0.0744, "step": 7195 }, { "epoch": 2.23, "learning_rate": 7.124624924984997e-06, "loss": 0.0772, "step": 7196 }, { "epoch": 2.23, "learning_rate": 7.124224844968994e-06, "loss": 0.0792, "step": 7197 }, { "epoch": 2.23, "learning_rate": 7.123824764952992e-06, "loss": 0.0412, "step": 7198 }, { "epoch": 2.23, "learning_rate": 7.123424684936988e-06, "loss": 0.1866, "step": 7199 }, { "epoch": 2.23, "learning_rate": 7.123024604920985e-06, "loss": 0.0483, "step": 7200 }, { "epoch": 2.24, "learning_rate": 7.122624524904982e-06, "loss": 0.0594, "step": 7201 }, { "epoch": 2.24, "learning_rate": 7.122224444888979e-06, "loss": 0.2649, "step": 7202 }, { "epoch": 2.24, "learning_rate": 7.121824364872976e-06, "loss": 0.1002, "step": 7203 }, { "epoch": 2.24, "learning_rate": 7.121424284856972e-06, "loss": 0.0324, "step": 7204 }, { "epoch": 2.24, "learning_rate": 7.121024204840968e-06, "loss": 0.087, "step": 7205 }, { "epoch": 2.24, "learning_rate": 7.120624124824965e-06, "loss": 0.1034, "step": 7206 }, { "epoch": 2.24, "learning_rate": 7.120224044808962e-06, "loss": 0.0838, "step": 7207 }, { "epoch": 2.24, "learning_rate": 7.119823964792959e-06, "loss": 0.0629, "step": 7208 }, { "epoch": 2.24, "learning_rate": 7.119423884776956e-06, "loss": 0.1182, "step": 7209 }, { "epoch": 2.24, "learning_rate": 7.119023804760953e-06, "loss": 0.1661, "step": 7210 }, { "epoch": 2.24, "learning_rate": 7.118623724744949e-06, "loss": 0.0459, "step": 7211 }, { "epoch": 2.24, "learning_rate": 7.118223644728947e-06, "loss": 0.0994, "step": 7212 }, { "epoch": 2.24, "learning_rate": 7.117823564712944e-06, "loss": 0.073, "step": 7213 }, { "epoch": 2.24, "learning_rate": 7.11742348469694e-06, "loss": 0.0896, "step": 7214 }, { "epoch": 2.24, "learning_rate": 7.117023404680937e-06, "loss": 0.0956, "step": 7215 }, { "epoch": 2.24, "learning_rate": 7.1166233246649345e-06, "loss": 0.0556, "step": 7216 }, { "epoch": 2.24, "learning_rate": 7.1162232446489295e-06, "loss": 0.0923, "step": 7217 }, { "epoch": 2.24, "learning_rate": 7.115823164632927e-06, "loss": 0.0974, "step": 7218 }, { "epoch": 2.24, "learning_rate": 7.115423084616924e-06, "loss": 0.0971, "step": 7219 }, { "epoch": 2.24, "learning_rate": 7.11502300460092e-06, "loss": 0.088, "step": 7220 }, { "epoch": 2.24, "learning_rate": 7.114622924584917e-06, "loss": 0.06, "step": 7221 }, { "epoch": 2.24, "learning_rate": 7.114222844568915e-06, "loss": 0.0755, "step": 7222 }, { "epoch": 2.24, "learning_rate": 7.113822764552911e-06, "loss": 0.0935, "step": 7223 }, { "epoch": 2.24, "learning_rate": 7.113422684536908e-06, "loss": 0.07, "step": 7224 }, { "epoch": 2.24, "learning_rate": 7.113022604520905e-06, "loss": 0.0563, "step": 7225 }, { "epoch": 2.24, "learning_rate": 7.112622524504902e-06, "loss": 0.234, "step": 7226 }, { "epoch": 2.24, "learning_rate": 7.112222444488899e-06, "loss": 0.0865, "step": 7227 }, { "epoch": 2.24, "learning_rate": 7.111822364472896e-06, "loss": 0.1386, "step": 7228 }, { "epoch": 2.24, "learning_rate": 7.1114222844568914e-06, "loss": 0.123, "step": 7229 }, { "epoch": 2.24, "learning_rate": 7.111022204440888e-06, "loss": 0.1651, "step": 7230 }, { "epoch": 2.24, "learning_rate": 7.110622124424885e-06, "loss": 0.0694, "step": 7231 }, { "epoch": 2.24, "learning_rate": 7.110222044408882e-06, "loss": 0.0608, "step": 7232 }, { "epoch": 2.24, "learning_rate": 7.109821964392879e-06, "loss": 0.1034, "step": 7233 }, { "epoch": 2.25, "learning_rate": 7.109421884376876e-06, "loss": 0.0522, "step": 7234 }, { "epoch": 2.25, "learning_rate": 7.109021804360872e-06, "loss": 0.0481, "step": 7235 }, { "epoch": 2.25, "learning_rate": 7.10862172434487e-06, "loss": 0.0571, "step": 7236 }, { "epoch": 2.25, "learning_rate": 7.108221644328867e-06, "loss": 0.086, "step": 7237 }, { "epoch": 2.25, "learning_rate": 7.107821564312863e-06, "loss": 0.1418, "step": 7238 }, { "epoch": 2.25, "learning_rate": 7.10742148429686e-06, "loss": 0.0975, "step": 7239 }, { "epoch": 2.25, "learning_rate": 7.1070214042808575e-06, "loss": 0.0514, "step": 7240 }, { "epoch": 2.25, "learning_rate": 7.106621324264854e-06, "loss": 0.0815, "step": 7241 }, { "epoch": 2.25, "learning_rate": 7.10622124424885e-06, "loss": 0.1771, "step": 7242 }, { "epoch": 2.25, "learning_rate": 7.105821164232847e-06, "loss": 0.0393, "step": 7243 }, { "epoch": 2.25, "learning_rate": 7.105421084216843e-06, "loss": 0.1592, "step": 7244 }, { "epoch": 2.25, "learning_rate": 7.10502100420084e-06, "loss": 0.1031, "step": 7245 }, { "epoch": 2.25, "learning_rate": 7.104620924184838e-06, "loss": 0.1508, "step": 7246 }, { "epoch": 2.25, "learning_rate": 7.104220844168834e-06, "loss": 0.0534, "step": 7247 }, { "epoch": 2.25, "learning_rate": 7.103820764152831e-06, "loss": 0.0712, "step": 7248 }, { "epoch": 2.25, "learning_rate": 7.103420684136828e-06, "loss": 0.0971, "step": 7249 }, { "epoch": 2.25, "learning_rate": 7.103020604120825e-06, "loss": 0.0839, "step": 7250 }, { "epoch": 2.25, "learning_rate": 7.102620524104822e-06, "loss": 0.1788, "step": 7251 }, { "epoch": 2.25, "learning_rate": 7.102220444088819e-06, "loss": 0.1138, "step": 7252 }, { "epoch": 2.25, "learning_rate": 7.101820364072815e-06, "loss": 0.0414, "step": 7253 }, { "epoch": 2.25, "learning_rate": 7.101420284056813e-06, "loss": 0.087, "step": 7254 }, { "epoch": 2.25, "learning_rate": 7.101020204040809e-06, "loss": 0.0747, "step": 7255 }, { "epoch": 2.25, "learning_rate": 7.100620124024805e-06, "loss": 0.0795, "step": 7256 }, { "epoch": 2.25, "learning_rate": 7.100220044008802e-06, "loss": 0.0798, "step": 7257 }, { "epoch": 2.25, "learning_rate": 7.099819963992799e-06, "loss": 0.098, "step": 7258 }, { "epoch": 2.25, "learning_rate": 7.099419883976795e-06, "loss": 0.1066, "step": 7259 }, { "epoch": 2.25, "learning_rate": 7.099019803960793e-06, "loss": 0.1386, "step": 7260 }, { "epoch": 2.25, "learning_rate": 7.09861972394479e-06, "loss": 0.0973, "step": 7261 }, { "epoch": 2.25, "learning_rate": 7.098219643928786e-06, "loss": 0.0718, "step": 7262 }, { "epoch": 2.25, "learning_rate": 7.097819563912783e-06, "loss": 0.0991, "step": 7263 }, { "epoch": 2.25, "learning_rate": 7.0974194838967805e-06, "loss": 0.0612, "step": 7264 }, { "epoch": 2.25, "learning_rate": 7.097019403880777e-06, "loss": 0.1021, "step": 7265 }, { "epoch": 2.26, "learning_rate": 7.096619323864774e-06, "loss": 0.0729, "step": 7266 }, { "epoch": 2.26, "learning_rate": 7.09621924384877e-06, "loss": 0.0496, "step": 7267 }, { "epoch": 2.26, "learning_rate": 7.0958191638327664e-06, "loss": 0.0945, "step": 7268 }, { "epoch": 2.26, "learning_rate": 7.095419083816764e-06, "loss": 0.0373, "step": 7269 }, { "epoch": 2.26, "learning_rate": 7.095019003800761e-06, "loss": 0.0595, "step": 7270 }, { "epoch": 2.26, "learning_rate": 7.094618923784757e-06, "loss": 0.1406, "step": 7271 }, { "epoch": 2.26, "learning_rate": 7.094218843768754e-06, "loss": 0.0668, "step": 7272 }, { "epoch": 2.26, "learning_rate": 7.0938187637527516e-06, "loss": 0.324, "step": 7273 }, { "epoch": 2.26, "learning_rate": 7.093418683736748e-06, "loss": 0.0655, "step": 7274 }, { "epoch": 2.26, "learning_rate": 7.093018603720745e-06, "loss": 0.0711, "step": 7275 }, { "epoch": 2.26, "learning_rate": 7.092618523704742e-06, "loss": 0.132, "step": 7276 }, { "epoch": 2.26, "learning_rate": 7.092218443688738e-06, "loss": 0.07, "step": 7277 }, { "epoch": 2.26, "learning_rate": 7.091818363672736e-06, "loss": 0.0802, "step": 7278 }, { "epoch": 2.26, "learning_rate": 7.0914182836567325e-06, "loss": 0.089, "step": 7279 }, { "epoch": 2.26, "learning_rate": 7.091018203640728e-06, "loss": 0.0609, "step": 7280 }, { "epoch": 2.26, "learning_rate": 7.090618123624725e-06, "loss": 0.0705, "step": 7281 }, { "epoch": 2.26, "learning_rate": 7.090218043608722e-06, "loss": 0.104, "step": 7282 }, { "epoch": 2.26, "learning_rate": 7.089817963592719e-06, "loss": 0.079, "step": 7283 }, { "epoch": 2.26, "learning_rate": 7.089417883576716e-06, "loss": 0.1102, "step": 7284 }, { "epoch": 2.26, "learning_rate": 7.089017803560713e-06, "loss": 0.0948, "step": 7285 }, { "epoch": 2.26, "learning_rate": 7.088617723544709e-06, "loss": 0.0833, "step": 7286 }, { "epoch": 2.26, "learning_rate": 7.088217643528707e-06, "loss": 0.0674, "step": 7287 }, { "epoch": 2.26, "learning_rate": 7.0878175635127035e-06, "loss": 0.0532, "step": 7288 }, { "epoch": 2.26, "learning_rate": 7.0874174834967e-06, "loss": 0.1427, "step": 7289 }, { "epoch": 2.26, "learning_rate": 7.087017403480697e-06, "loss": 0.0475, "step": 7290 }, { "epoch": 2.26, "learning_rate": 7.086617323464694e-06, "loss": 0.0885, "step": 7291 }, { "epoch": 2.26, "learning_rate": 7.0862172434486894e-06, "loss": 0.1058, "step": 7292 }, { "epoch": 2.26, "learning_rate": 7.085817163432687e-06, "loss": 0.0511, "step": 7293 }, { "epoch": 2.26, "learning_rate": 7.085417083416684e-06, "loss": 0.0843, "step": 7294 }, { "epoch": 2.26, "learning_rate": 7.08501700340068e-06, "loss": 0.2872, "step": 7295 }, { "epoch": 2.26, "learning_rate": 7.084616923384677e-06, "loss": 0.0577, "step": 7296 }, { "epoch": 2.26, "learning_rate": 7.0842168433686746e-06, "loss": 0.0924, "step": 7297 }, { "epoch": 2.27, "learning_rate": 7.083816763352671e-06, "loss": 0.0799, "step": 7298 }, { "epoch": 2.27, "learning_rate": 7.083416683336668e-06, "loss": 0.0532, "step": 7299 }, { "epoch": 2.27, "learning_rate": 7.083016603320665e-06, "loss": 0.1406, "step": 7300 }, { "epoch": 2.27, "learning_rate": 7.082616523304662e-06, "loss": 0.1666, "step": 7301 }, { "epoch": 2.27, "learning_rate": 7.082216443288659e-06, "loss": 0.0588, "step": 7302 }, { "epoch": 2.27, "learning_rate": 7.0818163632726555e-06, "loss": 0.0399, "step": 7303 }, { "epoch": 2.27, "learning_rate": 7.081416283256652e-06, "loss": 0.0646, "step": 7304 }, { "epoch": 2.27, "learning_rate": 7.081016203240648e-06, "loss": 0.1492, "step": 7305 }, { "epoch": 2.27, "learning_rate": 7.080616123224645e-06, "loss": 0.0585, "step": 7306 }, { "epoch": 2.27, "learning_rate": 7.080216043208642e-06, "loss": 0.1258, "step": 7307 }, { "epoch": 2.27, "learning_rate": 7.079815963192639e-06, "loss": 0.0805, "step": 7308 }, { "epoch": 2.27, "learning_rate": 7.079415883176636e-06, "loss": 0.1266, "step": 7309 }, { "epoch": 2.27, "learning_rate": 7.079015803160632e-06, "loss": 0.0426, "step": 7310 }, { "epoch": 2.27, "learning_rate": 7.07861572314463e-06, "loss": 0.0742, "step": 7311 }, { "epoch": 2.27, "learning_rate": 7.0782156431286266e-06, "loss": 0.0905, "step": 7312 }, { "epoch": 2.27, "learning_rate": 7.077815563112623e-06, "loss": 0.0949, "step": 7313 }, { "epoch": 2.27, "learning_rate": 7.07741548309662e-06, "loss": 0.0823, "step": 7314 }, { "epoch": 2.27, "learning_rate": 7.0770154030806175e-06, "loss": 0.0998, "step": 7315 }, { "epoch": 2.27, "learning_rate": 7.076615323064614e-06, "loss": 0.1101, "step": 7316 }, { "epoch": 2.27, "learning_rate": 7.076215243048611e-06, "loss": 0.0277, "step": 7317 }, { "epoch": 2.27, "learning_rate": 7.075815163032607e-06, "loss": 0.0653, "step": 7318 }, { "epoch": 2.27, "learning_rate": 7.075415083016603e-06, "loss": 0.0555, "step": 7319 }, { "epoch": 2.27, "learning_rate": 7.0750150030006e-06, "loss": 0.0758, "step": 7320 }, { "epoch": 2.27, "learning_rate": 7.0746149229845976e-06, "loss": 0.0983, "step": 7321 }, { "epoch": 2.27, "learning_rate": 7.074214842968594e-06, "loss": 0.0508, "step": 7322 }, { "epoch": 2.27, "learning_rate": 7.073814762952591e-06, "loss": 0.0818, "step": 7323 }, { "epoch": 2.27, "learning_rate": 7.073414682936588e-06, "loss": 0.0793, "step": 7324 }, { "epoch": 2.27, "learning_rate": 7.073014602920585e-06, "loss": 0.0413, "step": 7325 }, { "epoch": 2.27, "learning_rate": 7.072614522904582e-06, "loss": 0.0695, "step": 7326 }, { "epoch": 2.27, "learning_rate": 7.0722144428885785e-06, "loss": 0.0749, "step": 7327 }, { "epoch": 2.27, "learning_rate": 7.071814362872575e-06, "loss": 0.0992, "step": 7328 }, { "epoch": 2.27, "learning_rate": 7.071414282856573e-06, "loss": 0.0984, "step": 7329 }, { "epoch": 2.28, "learning_rate": 7.071014202840568e-06, "loss": 0.0977, "step": 7330 }, { "epoch": 2.28, "learning_rate": 7.070614122824565e-06, "loss": 0.0738, "step": 7331 }, { "epoch": 2.28, "learning_rate": 7.070214042808562e-06, "loss": 0.043, "step": 7332 }, { "epoch": 2.28, "learning_rate": 7.069813962792559e-06, "loss": 0.095, "step": 7333 }, { "epoch": 2.28, "learning_rate": 7.069413882776555e-06, "loss": 0.1183, "step": 7334 }, { "epoch": 2.28, "learning_rate": 7.069013802760553e-06, "loss": 0.0734, "step": 7335 }, { "epoch": 2.28, "learning_rate": 7.0686137227445496e-06, "loss": 0.0704, "step": 7336 }, { "epoch": 2.28, "learning_rate": 7.068213642728546e-06, "loss": 0.0298, "step": 7337 }, { "epoch": 2.28, "learning_rate": 7.067813562712543e-06, "loss": 0.1885, "step": 7338 }, { "epoch": 2.28, "learning_rate": 7.0674134826965405e-06, "loss": 0.0834, "step": 7339 }, { "epoch": 2.28, "learning_rate": 7.067013402680537e-06, "loss": 0.0743, "step": 7340 }, { "epoch": 2.28, "learning_rate": 7.066613322664534e-06, "loss": 0.0622, "step": 7341 }, { "epoch": 2.28, "learning_rate": 7.0662132426485305e-06, "loss": 0.0652, "step": 7342 }, { "epoch": 2.28, "learning_rate": 7.065813162632526e-06, "loss": 0.037, "step": 7343 }, { "epoch": 2.28, "learning_rate": 7.065413082616523e-06, "loss": 0.2176, "step": 7344 }, { "epoch": 2.28, "learning_rate": 7.065013002600521e-06, "loss": 0.0632, "step": 7345 }, { "epoch": 2.28, "learning_rate": 7.064612922584517e-06, "loss": 0.0589, "step": 7346 }, { "epoch": 2.28, "learning_rate": 7.064212842568514e-06, "loss": 0.0609, "step": 7347 }, { "epoch": 2.28, "learning_rate": 7.063812762552511e-06, "loss": 0.063, "step": 7348 }, { "epoch": 2.28, "learning_rate": 7.063412682536508e-06, "loss": 0.048, "step": 7349 }, { "epoch": 2.28, "learning_rate": 7.063012602520505e-06, "loss": 0.1576, "step": 7350 }, { "epoch": 2.28, "learning_rate": 7.0626125225045016e-06, "loss": 0.086, "step": 7351 }, { "epoch": 2.28, "learning_rate": 7.062212442488498e-06, "loss": 0.1266, "step": 7352 }, { "epoch": 2.28, "learning_rate": 7.061812362472496e-06, "loss": 0.0273, "step": 7353 }, { "epoch": 2.28, "learning_rate": 7.0614122824564925e-06, "loss": 0.0832, "step": 7354 }, { "epoch": 2.28, "learning_rate": 7.061012202440488e-06, "loss": 0.0928, "step": 7355 }, { "epoch": 2.28, "learning_rate": 7.060612122424485e-06, "loss": 0.0567, "step": 7356 }, { "epoch": 2.28, "learning_rate": 7.060212042408482e-06, "loss": 0.1247, "step": 7357 }, { "epoch": 2.28, "learning_rate": 7.059811962392479e-06, "loss": 0.0777, "step": 7358 }, { "epoch": 2.28, "learning_rate": 7.059411882376476e-06, "loss": 0.1454, "step": 7359 }, { "epoch": 2.28, "learning_rate": 7.0590118023604726e-06, "loss": 0.0781, "step": 7360 }, { "epoch": 2.28, "learning_rate": 7.058611722344469e-06, "loss": 0.0544, "step": 7361 }, { "epoch": 2.29, "learning_rate": 7.058211642328466e-06, "loss": 0.0916, "step": 7362 }, { "epoch": 2.29, "learning_rate": 7.0578115623124635e-06, "loss": 0.0154, "step": 7363 }, { "epoch": 2.29, "learning_rate": 7.05741148229646e-06, "loss": 0.0812, "step": 7364 }, { "epoch": 2.29, "learning_rate": 7.057011402280457e-06, "loss": 0.057, "step": 7365 }, { "epoch": 2.29, "learning_rate": 7.0566113222644535e-06, "loss": 0.1559, "step": 7366 }, { "epoch": 2.29, "learning_rate": 7.056211242248451e-06, "loss": 0.1704, "step": 7367 }, { "epoch": 2.29, "learning_rate": 7.055811162232447e-06, "loss": 0.1019, "step": 7368 }, { "epoch": 2.29, "learning_rate": 7.055411082216444e-06, "loss": 0.0737, "step": 7369 }, { "epoch": 2.29, "learning_rate": 7.05501100220044e-06, "loss": 0.1015, "step": 7370 }, { "epoch": 2.29, "learning_rate": 7.054610922184437e-06, "loss": 0.0435, "step": 7371 }, { "epoch": 2.29, "learning_rate": 7.0542108421684345e-06, "loss": 0.1133, "step": 7372 }, { "epoch": 2.29, "learning_rate": 7.053810762152431e-06, "loss": 0.0896, "step": 7373 }, { "epoch": 2.29, "learning_rate": 7.053410682136428e-06, "loss": 0.0337, "step": 7374 }, { "epoch": 2.29, "learning_rate": 7.0530106021204246e-06, "loss": 0.0508, "step": 7375 }, { "epoch": 2.29, "learning_rate": 7.052610522104421e-06, "loss": 0.0547, "step": 7376 }, { "epoch": 2.29, "learning_rate": 7.052210442088419e-06, "loss": 0.2354, "step": 7377 }, { "epoch": 2.29, "learning_rate": 7.0518103620724155e-06, "loss": 0.1102, "step": 7378 }, { "epoch": 2.29, "learning_rate": 7.051410282056412e-06, "loss": 0.0631, "step": 7379 }, { "epoch": 2.29, "learning_rate": 7.051010202040409e-06, "loss": 0.09, "step": 7380 }, { "epoch": 2.29, "learning_rate": 7.050610122024405e-06, "loss": 0.1519, "step": 7381 }, { "epoch": 2.29, "learning_rate": 7.050210042008402e-06, "loss": 0.06, "step": 7382 }, { "epoch": 2.29, "learning_rate": 7.049809961992399e-06, "loss": 0.1026, "step": 7383 }, { "epoch": 2.29, "learning_rate": 7.049409881976396e-06, "loss": 0.0355, "step": 7384 }, { "epoch": 2.29, "learning_rate": 7.049009801960392e-06, "loss": 0.0507, "step": 7385 }, { "epoch": 2.29, "learning_rate": 7.04860972194439e-06, "loss": 0.0806, "step": 7386 }, { "epoch": 2.29, "learning_rate": 7.0482096419283865e-06, "loss": 0.0999, "step": 7387 }, { "epoch": 2.29, "learning_rate": 7.047809561912383e-06, "loss": 0.0511, "step": 7388 }, { "epoch": 2.29, "learning_rate": 7.04740948189638e-06, "loss": 0.068, "step": 7389 }, { "epoch": 2.29, "learning_rate": 7.0470094018803766e-06, "loss": 0.1112, "step": 7390 }, { "epoch": 2.29, "learning_rate": 7.046609321864374e-06, "loss": 0.0591, "step": 7391 }, { "epoch": 2.29, "learning_rate": 7.046209241848371e-06, "loss": 0.0926, "step": 7392 }, { "epoch": 2.29, "learning_rate": 7.045809161832367e-06, "loss": 0.0942, "step": 7393 }, { "epoch": 2.29, "learning_rate": 7.045409081816363e-06, "loss": 0.2108, "step": 7394 }, { "epoch": 2.3, "learning_rate": 7.04500900180036e-06, "loss": 0.0923, "step": 7395 }, { "epoch": 2.3, "learning_rate": 7.0446089217843575e-06, "loss": 0.0427, "step": 7396 }, { "epoch": 2.3, "learning_rate": 7.044208841768354e-06, "loss": 0.1014, "step": 7397 }, { "epoch": 2.3, "learning_rate": 7.043808761752351e-06, "loss": 0.1745, "step": 7398 }, { "epoch": 2.3, "learning_rate": 7.0434086817363476e-06, "loss": 0.0524, "step": 7399 }, { "epoch": 2.3, "learning_rate": 7.043008601720345e-06, "loss": 0.1374, "step": 7400 }, { "epoch": 2.3, "learning_rate": 7.042608521704342e-06, "loss": 0.1099, "step": 7401 }, { "epoch": 2.3, "learning_rate": 7.0422084416883385e-06, "loss": 0.0412, "step": 7402 }, { "epoch": 2.3, "learning_rate": 7.041808361672335e-06, "loss": 0.1055, "step": 7403 }, { "epoch": 2.3, "learning_rate": 7.041408281656333e-06, "loss": 0.119, "step": 7404 }, { "epoch": 2.3, "learning_rate": 7.041008201640329e-06, "loss": 0.0938, "step": 7405 }, { "epoch": 2.3, "learning_rate": 7.040608121624325e-06, "loss": 0.0607, "step": 7406 }, { "epoch": 2.3, "learning_rate": 7.040208041608322e-06, "loss": 0.1735, "step": 7407 }, { "epoch": 2.3, "learning_rate": 7.039807961592319e-06, "loss": 0.091, "step": 7408 }, { "epoch": 2.3, "learning_rate": 7.039407881576315e-06, "loss": 0.0851, "step": 7409 }, { "epoch": 2.3, "learning_rate": 7.039007801560313e-06, "loss": 0.0816, "step": 7410 }, { "epoch": 2.3, "learning_rate": 7.0386077215443095e-06, "loss": 0.0942, "step": 7411 }, { "epoch": 2.3, "learning_rate": 7.038207641528306e-06, "loss": 0.0746, "step": 7412 }, { "epoch": 2.3, "learning_rate": 7.037807561512303e-06, "loss": 0.083, "step": 7413 }, { "epoch": 2.3, "learning_rate": 7.0374074814963e-06, "loss": 0.1352, "step": 7414 }, { "epoch": 2.3, "learning_rate": 7.037007401480297e-06, "loss": 0.1322, "step": 7415 }, { "epoch": 2.3, "learning_rate": 7.036607321464294e-06, "loss": 0.1153, "step": 7416 }, { "epoch": 2.3, "learning_rate": 7.0362072414482905e-06, "loss": 0.0849, "step": 7417 }, { "epoch": 2.3, "learning_rate": 7.035807161432286e-06, "loss": 0.06, "step": 7418 }, { "epoch": 2.3, "learning_rate": 7.035407081416283e-06, "loss": 0.085, "step": 7419 }, { "epoch": 2.3, "learning_rate": 7.0350070014002805e-06, "loss": 0.1305, "step": 7420 }, { "epoch": 2.3, "learning_rate": 7.034606921384277e-06, "loss": 0.0765, "step": 7421 }, { "epoch": 2.3, "learning_rate": 7.034206841368274e-06, "loss": 0.0781, "step": 7422 }, { "epoch": 2.3, "learning_rate": 7.033806761352271e-06, "loss": 0.0365, "step": 7423 }, { "epoch": 2.3, "learning_rate": 7.033406681336268e-06, "loss": 0.0852, "step": 7424 }, { "epoch": 2.3, "learning_rate": 7.033006601320265e-06, "loss": 0.0627, "step": 7425 }, { "epoch": 2.3, "learning_rate": 7.0326065213042615e-06, "loss": 0.1098, "step": 7426 }, { "epoch": 2.31, "learning_rate": 7.032206441288258e-06, "loss": 0.1116, "step": 7427 }, { "epoch": 2.31, "learning_rate": 7.031806361272256e-06, "loss": 0.1445, "step": 7428 }, { "epoch": 2.31, "learning_rate": 7.031406281256252e-06, "loss": 0.0973, "step": 7429 }, { "epoch": 2.31, "learning_rate": 7.031006201240249e-06, "loss": 0.0314, "step": 7430 }, { "epoch": 2.31, "learning_rate": 7.030606121224245e-06, "loss": 0.2396, "step": 7431 }, { "epoch": 2.31, "learning_rate": 7.030206041208242e-06, "loss": 0.0612, "step": 7432 }, { "epoch": 2.31, "learning_rate": 7.029805961192238e-06, "loss": 0.1245, "step": 7433 }, { "epoch": 2.31, "learning_rate": 7.029405881176236e-06, "loss": 0.1106, "step": 7434 }, { "epoch": 2.31, "learning_rate": 7.0290058011602325e-06, "loss": 0.0517, "step": 7435 }, { "epoch": 2.31, "learning_rate": 7.028605721144229e-06, "loss": 0.0905, "step": 7436 }, { "epoch": 2.31, "learning_rate": 7.028205641128226e-06, "loss": 0.0439, "step": 7437 }, { "epoch": 2.31, "learning_rate": 7.027805561112223e-06, "loss": 0.101, "step": 7438 }, { "epoch": 2.31, "learning_rate": 7.02740548109622e-06, "loss": 0.099, "step": 7439 }, { "epoch": 2.31, "learning_rate": 7.027005401080217e-06, "loss": 0.1793, "step": 7440 }, { "epoch": 2.31, "learning_rate": 7.0266053210642135e-06, "loss": 0.0629, "step": 7441 }, { "epoch": 2.31, "learning_rate": 7.026205241048211e-06, "loss": 0.0573, "step": 7442 }, { "epoch": 2.31, "learning_rate": 7.025805161032208e-06, "loss": 0.2158, "step": 7443 }, { "epoch": 2.31, "learning_rate": 7.0254050810162035e-06, "loss": 0.118, "step": 7444 }, { "epoch": 2.31, "learning_rate": 7.0250050010002e-06, "loss": 0.0666, "step": 7445 }, { "epoch": 2.31, "learning_rate": 7.024604920984197e-06, "loss": 0.1218, "step": 7446 }, { "epoch": 2.31, "learning_rate": 7.024204840968194e-06, "loss": 0.071, "step": 7447 }, { "epoch": 2.31, "learning_rate": 7.023804760952191e-06, "loss": 0.0604, "step": 7448 }, { "epoch": 2.31, "learning_rate": 7.023404680936188e-06, "loss": 0.0847, "step": 7449 }, { "epoch": 2.31, "learning_rate": 7.0230046009201845e-06, "loss": 0.1539, "step": 7450 }, { "epoch": 2.31, "learning_rate": 7.022604520904181e-06, "loss": 0.0685, "step": 7451 }, { "epoch": 2.31, "learning_rate": 7.022204440888179e-06, "loss": 0.0272, "step": 7452 }, { "epoch": 2.31, "learning_rate": 7.021804360872175e-06, "loss": 0.0594, "step": 7453 }, { "epoch": 2.31, "learning_rate": 7.021404280856172e-06, "loss": 0.1642, "step": 7454 }, { "epoch": 2.31, "learning_rate": 7.021004200840169e-06, "loss": 0.0567, "step": 7455 }, { "epoch": 2.31, "learning_rate": 7.020604120824165e-06, "loss": 0.0932, "step": 7456 }, { "epoch": 2.31, "learning_rate": 7.020204040808162e-06, "loss": 0.0843, "step": 7457 }, { "epoch": 2.31, "learning_rate": 7.019803960792159e-06, "loss": 0.1014, "step": 7458 }, { "epoch": 2.32, "learning_rate": 7.0194038807761555e-06, "loss": 0.0663, "step": 7459 }, { "epoch": 2.32, "learning_rate": 7.019003800760152e-06, "loss": 0.0583, "step": 7460 }, { "epoch": 2.32, "learning_rate": 7.018603720744149e-06, "loss": 0.0794, "step": 7461 }, { "epoch": 2.32, "learning_rate": 7.0182036407281464e-06, "loss": 0.144, "step": 7462 }, { "epoch": 2.32, "learning_rate": 7.017803560712143e-06, "loss": 0.0458, "step": 7463 }, { "epoch": 2.32, "learning_rate": 7.01740348069614e-06, "loss": 0.0519, "step": 7464 }, { "epoch": 2.32, "learning_rate": 7.0170034006801365e-06, "loss": 0.0878, "step": 7465 }, { "epoch": 2.32, "learning_rate": 7.016603320664134e-06, "loss": 0.1374, "step": 7466 }, { "epoch": 2.32, "learning_rate": 7.016203240648131e-06, "loss": 0.1012, "step": 7467 }, { "epoch": 2.32, "learning_rate": 7.015803160632127e-06, "loss": 0.033, "step": 7468 }, { "epoch": 2.32, "learning_rate": 7.015403080616123e-06, "loss": 0.0476, "step": 7469 }, { "epoch": 2.32, "learning_rate": 7.01500300060012e-06, "loss": 0.1075, "step": 7470 }, { "epoch": 2.32, "learning_rate": 7.0146029205841175e-06, "loss": 0.0482, "step": 7471 }, { "epoch": 2.32, "learning_rate": 7.014202840568114e-06, "loss": 0.0369, "step": 7472 }, { "epoch": 2.32, "learning_rate": 7.013802760552111e-06, "loss": 0.1884, "step": 7473 }, { "epoch": 2.32, "learning_rate": 7.0134026805361075e-06, "loss": 0.083, "step": 7474 }, { "epoch": 2.32, "learning_rate": 7.013002600520104e-06, "loss": 0.1307, "step": 7475 }, { "epoch": 2.32, "learning_rate": 7.012602520504102e-06, "loss": 0.1845, "step": 7476 }, { "epoch": 2.32, "learning_rate": 7.012202440488098e-06, "loss": 0.1043, "step": 7477 }, { "epoch": 2.32, "learning_rate": 7.011802360472095e-06, "loss": 0.0747, "step": 7478 }, { "epoch": 2.32, "learning_rate": 7.011402280456092e-06, "loss": 0.0943, "step": 7479 }, { "epoch": 2.32, "learning_rate": 7.011002200440089e-06, "loss": 0.1206, "step": 7480 }, { "epoch": 2.32, "learning_rate": 7.010602120424086e-06, "loss": 0.0698, "step": 7481 }, { "epoch": 2.32, "learning_rate": 7.010202040408082e-06, "loss": 0.0833, "step": 7482 }, { "epoch": 2.32, "learning_rate": 7.0098019603920785e-06, "loss": 0.2718, "step": 7483 }, { "epoch": 2.32, "learning_rate": 7.009401880376075e-06, "loss": 0.1049, "step": 7484 }, { "epoch": 2.32, "learning_rate": 7.009001800360073e-06, "loss": 0.0741, "step": 7485 }, { "epoch": 2.32, "learning_rate": 7.0086017203440694e-06, "loss": 0.0466, "step": 7486 }, { "epoch": 2.32, "learning_rate": 7.008201640328066e-06, "loss": 0.0802, "step": 7487 }, { "epoch": 2.32, "learning_rate": 7.007801560312063e-06, "loss": 0.0449, "step": 7488 }, { "epoch": 2.32, "learning_rate": 7.00740148029606e-06, "loss": 0.1071, "step": 7489 }, { "epoch": 2.32, "learning_rate": 7.007001400280057e-06, "loss": 0.106, "step": 7490 }, { "epoch": 2.33, "learning_rate": 7.006601320264054e-06, "loss": 0.1895, "step": 7491 }, { "epoch": 2.33, "learning_rate": 7.00620124024805e-06, "loss": 0.0984, "step": 7492 }, { "epoch": 2.33, "learning_rate": 7.005801160232047e-06, "loss": 0.1275, "step": 7493 }, { "epoch": 2.33, "learning_rate": 7.005401080216043e-06, "loss": 0.1867, "step": 7494 }, { "epoch": 2.33, "learning_rate": 7.0050010002000405e-06, "loss": 0.0778, "step": 7495 }, { "epoch": 2.33, "learning_rate": 7.004600920184037e-06, "loss": 0.0729, "step": 7496 }, { "epoch": 2.33, "learning_rate": 7.004200840168034e-06, "loss": 0.0502, "step": 7497 }, { "epoch": 2.33, "learning_rate": 7.0038007601520305e-06, "loss": 0.1219, "step": 7498 }, { "epoch": 2.33, "learning_rate": 7.003400680136028e-06, "loss": 0.0838, "step": 7499 }, { "epoch": 2.33, "learning_rate": 7.003000600120025e-06, "loss": 0.1104, "step": 7500 }, { "epoch": 2.33, "learning_rate": 7.0026005201040214e-06, "loss": 0.0553, "step": 7501 }, { "epoch": 2.33, "learning_rate": 7.002200440088018e-06, "loss": 0.0606, "step": 7502 }, { "epoch": 2.33, "learning_rate": 7.001800360072016e-06, "loss": 0.0722, "step": 7503 }, { "epoch": 2.33, "learning_rate": 7.001400280056012e-06, "loss": 0.0663, "step": 7504 }, { "epoch": 2.33, "learning_rate": 7.001000200040009e-06, "loss": 0.116, "step": 7505 }, { "epoch": 2.33, "learning_rate": 7.000600120024006e-06, "loss": 0.0536, "step": 7506 }, { "epoch": 2.33, "learning_rate": 7.0002000400080016e-06, "loss": 0.0505, "step": 7507 }, { "epoch": 2.33, "learning_rate": 6.999799959991998e-06, "loss": 0.2206, "step": 7508 }, { "epoch": 2.33, "learning_rate": 6.999399879975996e-06, "loss": 0.0595, "step": 7509 }, { "epoch": 2.33, "learning_rate": 6.9989997999599925e-06, "loss": 0.0894, "step": 7510 }, { "epoch": 2.33, "learning_rate": 6.998599719943989e-06, "loss": 0.1085, "step": 7511 }, { "epoch": 2.33, "learning_rate": 6.998199639927986e-06, "loss": 0.157, "step": 7512 }, { "epoch": 2.33, "learning_rate": 6.997799559911983e-06, "loss": 0.125, "step": 7513 }, { "epoch": 2.33, "learning_rate": 6.99739947989598e-06, "loss": 0.0514, "step": 7514 }, { "epoch": 2.33, "learning_rate": 6.996999399879977e-06, "loss": 0.1526, "step": 7515 }, { "epoch": 2.33, "learning_rate": 6.996599319863973e-06, "loss": 0.062, "step": 7516 }, { "epoch": 2.33, "learning_rate": 6.996199239847971e-06, "loss": 0.0479, "step": 7517 }, { "epoch": 2.33, "learning_rate": 6.995799159831968e-06, "loss": 0.124, "step": 7518 }, { "epoch": 2.33, "learning_rate": 6.9953990798159635e-06, "loss": 0.1327, "step": 7519 }, { "epoch": 2.33, "learning_rate": 6.99499899979996e-06, "loss": 0.0933, "step": 7520 }, { "epoch": 2.33, "learning_rate": 6.994598919783957e-06, "loss": 0.0977, "step": 7521 }, { "epoch": 2.33, "learning_rate": 6.9941988397679535e-06, "loss": 0.1619, "step": 7522 }, { "epoch": 2.33, "learning_rate": 6.993798759751951e-06, "loss": 0.0747, "step": 7523 }, { "epoch": 2.34, "learning_rate": 6.993398679735948e-06, "loss": 0.1401, "step": 7524 }, { "epoch": 2.34, "learning_rate": 6.9929985997199444e-06, "loss": 0.1222, "step": 7525 }, { "epoch": 2.34, "learning_rate": 6.992598519703941e-06, "loss": 0.1818, "step": 7526 }, { "epoch": 2.34, "learning_rate": 6.992198439687939e-06, "loss": 0.0465, "step": 7527 }, { "epoch": 2.34, "learning_rate": 6.991798359671935e-06, "loss": 0.1823, "step": 7528 }, { "epoch": 2.34, "learning_rate": 6.991398279655932e-06, "loss": 0.0753, "step": 7529 }, { "epoch": 2.34, "learning_rate": 6.990998199639929e-06, "loss": 0.088, "step": 7530 }, { "epoch": 2.34, "learning_rate": 6.990598119623926e-06, "loss": 0.1234, "step": 7531 }, { "epoch": 2.34, "learning_rate": 6.990198039607921e-06, "loss": 0.1531, "step": 7532 }, { "epoch": 2.34, "learning_rate": 6.989797959591919e-06, "loss": 0.0525, "step": 7533 }, { "epoch": 2.34, "learning_rate": 6.9893978795759155e-06, "loss": 0.0568, "step": 7534 }, { "epoch": 2.34, "learning_rate": 6.988997799559912e-06, "loss": 0.1286, "step": 7535 }, { "epoch": 2.34, "learning_rate": 6.988597719543909e-06, "loss": 0.0857, "step": 7536 }, { "epoch": 2.34, "learning_rate": 6.988197639527906e-06, "loss": 0.0731, "step": 7537 }, { "epoch": 2.34, "learning_rate": 6.987797559511903e-06, "loss": 0.048, "step": 7538 }, { "epoch": 2.34, "learning_rate": 6.9873974794959e-06, "loss": 0.0864, "step": 7539 }, { "epoch": 2.34, "learning_rate": 6.9869973994798964e-06, "loss": 0.0768, "step": 7540 }, { "epoch": 2.34, "learning_rate": 6.986597319463894e-06, "loss": 0.042, "step": 7541 }, { "epoch": 2.34, "learning_rate": 6.986197239447891e-06, "loss": 0.0544, "step": 7542 }, { "epoch": 2.34, "learning_rate": 6.985797159431887e-06, "loss": 0.1492, "step": 7543 }, { "epoch": 2.34, "learning_rate": 6.985397079415884e-06, "loss": 0.0609, "step": 7544 }, { "epoch": 2.34, "learning_rate": 6.98499699939988e-06, "loss": 0.0756, "step": 7545 }, { "epoch": 2.34, "learning_rate": 6.9845969193838766e-06, "loss": 0.1054, "step": 7546 }, { "epoch": 2.34, "learning_rate": 6.984196839367874e-06, "loss": 0.1566, "step": 7547 }, { "epoch": 2.34, "learning_rate": 6.983796759351871e-06, "loss": 0.0613, "step": 7548 }, { "epoch": 2.34, "learning_rate": 6.9833966793358675e-06, "loss": 0.0666, "step": 7549 }, { "epoch": 2.34, "learning_rate": 6.982996599319864e-06, "loss": 0.0588, "step": 7550 }, { "epoch": 2.34, "learning_rate": 6.982596519303862e-06, "loss": 0.0911, "step": 7551 }, { "epoch": 2.34, "learning_rate": 6.982196439287858e-06, "loss": 0.0322, "step": 7552 }, { "epoch": 2.34, "learning_rate": 6.981796359271855e-06, "loss": 0.0798, "step": 7553 }, { "epoch": 2.34, "learning_rate": 6.981396279255852e-06, "loss": 0.0973, "step": 7554 }, { "epoch": 2.34, "learning_rate": 6.980996199239849e-06, "loss": 0.1622, "step": 7555 }, { "epoch": 2.35, "learning_rate": 6.980596119223846e-06, "loss": 0.1179, "step": 7556 }, { "epoch": 2.35, "learning_rate": 6.980196039207842e-06, "loss": 0.0834, "step": 7557 }, { "epoch": 2.35, "learning_rate": 6.9797959591918385e-06, "loss": 0.0863, "step": 7558 }, { "epoch": 2.35, "learning_rate": 6.979395879175835e-06, "loss": 0.307, "step": 7559 }, { "epoch": 2.35, "learning_rate": 6.978995799159832e-06, "loss": 0.0629, "step": 7560 }, { "epoch": 2.35, "learning_rate": 6.978595719143829e-06, "loss": 0.0402, "step": 7561 }, { "epoch": 2.35, "learning_rate": 6.978195639127826e-06, "loss": 0.1542, "step": 7562 }, { "epoch": 2.35, "learning_rate": 6.977795559111823e-06, "loss": 0.0758, "step": 7563 }, { "epoch": 2.35, "learning_rate": 6.9773954790958194e-06, "loss": 0.0257, "step": 7564 }, { "epoch": 2.35, "learning_rate": 6.976995399079817e-06, "loss": 0.108, "step": 7565 }, { "epoch": 2.35, "learning_rate": 6.976595319063814e-06, "loss": 0.0859, "step": 7566 }, { "epoch": 2.35, "learning_rate": 6.97619523904781e-06, "loss": 0.0737, "step": 7567 }, { "epoch": 2.35, "learning_rate": 6.975795159031807e-06, "loss": 0.1094, "step": 7568 }, { "epoch": 2.35, "learning_rate": 6.9753950790158046e-06, "loss": 0.0858, "step": 7569 }, { "epoch": 2.35, "learning_rate": 6.9749949989998e-06, "loss": 0.0956, "step": 7570 }, { "epoch": 2.35, "learning_rate": 6.974594918983797e-06, "loss": 0.0636, "step": 7571 }, { "epoch": 2.35, "learning_rate": 6.974194838967794e-06, "loss": 0.0444, "step": 7572 }, { "epoch": 2.35, "learning_rate": 6.9737947589517905e-06, "loss": 0.1021, "step": 7573 }, { "epoch": 2.35, "learning_rate": 6.973394678935788e-06, "loss": 0.1348, "step": 7574 }, { "epoch": 2.35, "learning_rate": 6.972994598919785e-06, "loss": 0.0594, "step": 7575 }, { "epoch": 2.35, "learning_rate": 6.972594518903781e-06, "loss": 0.112, "step": 7576 }, { "epoch": 2.35, "learning_rate": 6.972194438887778e-06, "loss": 0.1102, "step": 7577 }, { "epoch": 2.35, "learning_rate": 6.971794358871775e-06, "loss": 0.0282, "step": 7578 }, { "epoch": 2.35, "learning_rate": 6.971394278855772e-06, "loss": 0.0659, "step": 7579 }, { "epoch": 2.35, "learning_rate": 6.970994198839769e-06, "loss": 0.0581, "step": 7580 }, { "epoch": 2.35, "learning_rate": 6.970594118823766e-06, "loss": 0.0494, "step": 7581 }, { "epoch": 2.35, "learning_rate": 6.9701940388077615e-06, "loss": 0.07, "step": 7582 }, { "epoch": 2.35, "learning_rate": 6.969793958791758e-06, "loss": 0.0463, "step": 7583 }, { "epoch": 2.35, "learning_rate": 6.969393878775756e-06, "loss": 0.0791, "step": 7584 }, { "epoch": 2.35, "learning_rate": 6.968993798759752e-06, "loss": 0.0956, "step": 7585 }, { "epoch": 2.35, "learning_rate": 6.968593718743749e-06, "loss": 0.0829, "step": 7586 }, { "epoch": 2.35, "learning_rate": 6.968193638727746e-06, "loss": 0.0528, "step": 7587 }, { "epoch": 2.36, "learning_rate": 6.967793558711743e-06, "loss": 0.071, "step": 7588 }, { "epoch": 2.36, "learning_rate": 6.96739347869574e-06, "loss": 0.0749, "step": 7589 }, { "epoch": 2.36, "learning_rate": 6.966993398679737e-06, "loss": 0.2288, "step": 7590 }, { "epoch": 2.36, "learning_rate": 6.966593318663733e-06, "loss": 0.1468, "step": 7591 }, { "epoch": 2.36, "learning_rate": 6.96619323864773e-06, "loss": 0.0929, "step": 7592 }, { "epoch": 2.36, "learning_rate": 6.965793158631728e-06, "loss": 0.1199, "step": 7593 }, { "epoch": 2.36, "learning_rate": 6.965393078615724e-06, "loss": 0.0656, "step": 7594 }, { "epoch": 2.36, "learning_rate": 6.96499299859972e-06, "loss": 0.0757, "step": 7595 }, { "epoch": 2.36, "learning_rate": 6.964592918583717e-06, "loss": 0.0908, "step": 7596 }, { "epoch": 2.36, "learning_rate": 6.9641928385677135e-06, "loss": 0.111, "step": 7597 }, { "epoch": 2.36, "learning_rate": 6.963792758551711e-06, "loss": 0.0771, "step": 7598 }, { "epoch": 2.36, "learning_rate": 6.963392678535708e-06, "loss": 0.037, "step": 7599 }, { "epoch": 2.36, "learning_rate": 6.962992598519704e-06, "loss": 0.0689, "step": 7600 }, { "epoch": 2.36, "learning_rate": 6.962592518503701e-06, "loss": 0.0343, "step": 7601 }, { "epoch": 2.36, "learning_rate": 6.962192438487699e-06, "loss": 0.0981, "step": 7602 }, { "epoch": 2.36, "learning_rate": 6.961792358471695e-06, "loss": 0.208, "step": 7603 }, { "epoch": 2.36, "learning_rate": 6.961392278455692e-06, "loss": 0.0752, "step": 7604 }, { "epoch": 2.36, "learning_rate": 6.960992198439689e-06, "loss": 0.0418, "step": 7605 }, { "epoch": 2.36, "learning_rate": 6.960592118423685e-06, "loss": 0.1633, "step": 7606 }, { "epoch": 2.36, "learning_rate": 6.960192038407683e-06, "loss": 0.0998, "step": 7607 }, { "epoch": 2.36, "learning_rate": 6.959791958391679e-06, "loss": 0.1657, "step": 7608 }, { "epoch": 2.36, "learning_rate": 6.959391878375675e-06, "loss": 0.0809, "step": 7609 }, { "epoch": 2.36, "learning_rate": 6.958991798359672e-06, "loss": 0.2148, "step": 7610 }, { "epoch": 2.36, "learning_rate": 6.958591718343669e-06, "loss": 0.2045, "step": 7611 }, { "epoch": 2.36, "learning_rate": 6.958191638327666e-06, "loss": 0.097, "step": 7612 }, { "epoch": 2.36, "learning_rate": 6.957791558311663e-06, "loss": 0.0966, "step": 7613 }, { "epoch": 2.36, "learning_rate": 6.95739147829566e-06, "loss": 0.1601, "step": 7614 }, { "epoch": 2.36, "learning_rate": 6.956991398279656e-06, "loss": 0.0604, "step": 7615 }, { "epoch": 2.36, "learning_rate": 6.956591318263654e-06, "loss": 0.0736, "step": 7616 }, { "epoch": 2.36, "learning_rate": 6.956191238247651e-06, "loss": 0.0448, "step": 7617 }, { "epoch": 2.36, "learning_rate": 6.955791158231647e-06, "loss": 0.0841, "step": 7618 }, { "epoch": 2.36, "learning_rate": 6.955391078215644e-06, "loss": 0.1125, "step": 7619 }, { "epoch": 2.37, "learning_rate": 6.95499099819964e-06, "loss": 0.2097, "step": 7620 }, { "epoch": 2.37, "learning_rate": 6.9545909181836365e-06, "loss": 0.1555, "step": 7621 }, { "epoch": 2.37, "learning_rate": 6.954190838167634e-06, "loss": 0.1717, "step": 7622 }, { "epoch": 2.37, "learning_rate": 6.953790758151631e-06, "loss": 0.1521, "step": 7623 }, { "epoch": 2.37, "learning_rate": 6.953390678135627e-06, "loss": 0.0734, "step": 7624 }, { "epoch": 2.37, "learning_rate": 6.952990598119624e-06, "loss": 0.1692, "step": 7625 }, { "epoch": 2.37, "learning_rate": 6.952590518103622e-06, "loss": 0.0163, "step": 7626 }, { "epoch": 2.37, "learning_rate": 6.952190438087618e-06, "loss": 0.0477, "step": 7627 }, { "epoch": 2.37, "learning_rate": 6.951790358071615e-06, "loss": 0.081, "step": 7628 }, { "epoch": 2.37, "learning_rate": 6.951390278055612e-06, "loss": 0.0477, "step": 7629 }, { "epoch": 2.37, "learning_rate": 6.950990198039609e-06, "loss": 0.1027, "step": 7630 }, { "epoch": 2.37, "learning_rate": 6.950590118023606e-06, "loss": 0.084, "step": 7631 }, { "epoch": 2.37, "learning_rate": 6.950190038007603e-06, "loss": 0.2038, "step": 7632 }, { "epoch": 2.37, "learning_rate": 6.949789957991598e-06, "loss": 0.0687, "step": 7633 }, { "epoch": 2.37, "learning_rate": 6.949389877975595e-06, "loss": 0.0628, "step": 7634 }, { "epoch": 2.37, "learning_rate": 6.948989797959592e-06, "loss": 0.0543, "step": 7635 }, { "epoch": 2.37, "learning_rate": 6.948589717943589e-06, "loss": 0.1127, "step": 7636 }, { "epoch": 2.37, "learning_rate": 6.948189637927586e-06, "loss": 0.0696, "step": 7637 }, { "epoch": 2.37, "learning_rate": 6.947789557911583e-06, "loss": 0.0607, "step": 7638 }, { "epoch": 2.37, "learning_rate": 6.947389477895579e-06, "loss": 0.0634, "step": 7639 }, { "epoch": 2.37, "learning_rate": 6.946989397879577e-06, "loss": 0.0704, "step": 7640 }, { "epoch": 2.37, "learning_rate": 6.946589317863574e-06, "loss": 0.1498, "step": 7641 }, { "epoch": 2.37, "learning_rate": 6.94618923784757e-06, "loss": 0.1156, "step": 7642 }, { "epoch": 2.37, "learning_rate": 6.945789157831567e-06, "loss": 0.1171, "step": 7643 }, { "epoch": 2.37, "learning_rate": 6.9453890778155645e-06, "loss": 0.1053, "step": 7644 }, { "epoch": 2.37, "learning_rate": 6.94498899779956e-06, "loss": 0.0547, "step": 7645 }, { "epoch": 2.37, "learning_rate": 6.944588917783557e-06, "loss": 0.1149, "step": 7646 }, { "epoch": 2.37, "learning_rate": 6.944188837767554e-06, "loss": 0.0552, "step": 7647 }, { "epoch": 2.37, "learning_rate": 6.94378875775155e-06, "loss": 0.0589, "step": 7648 }, { "epoch": 2.37, "learning_rate": 6.943388677735547e-06, "loss": 0.1845, "step": 7649 }, { "epoch": 2.37, "learning_rate": 6.942988597719545e-06, "loss": 0.07, "step": 7650 }, { "epoch": 2.37, "learning_rate": 6.942588517703541e-06, "loss": 0.2091, "step": 7651 }, { "epoch": 2.38, "learning_rate": 6.942188437687538e-06, "loss": 0.1462, "step": 7652 }, { "epoch": 2.38, "learning_rate": 6.941788357671535e-06, "loss": 0.1223, "step": 7653 }, { "epoch": 2.38, "learning_rate": 6.941388277655532e-06, "loss": 0.201, "step": 7654 }, { "epoch": 2.38, "learning_rate": 6.940988197639529e-06, "loss": 0.0724, "step": 7655 }, { "epoch": 2.38, "learning_rate": 6.940588117623526e-06, "loss": 0.0571, "step": 7656 }, { "epoch": 2.38, "learning_rate": 6.940188037607522e-06, "loss": 0.0561, "step": 7657 }, { "epoch": 2.38, "learning_rate": 6.939787957591518e-06, "loss": 0.2249, "step": 7658 }, { "epoch": 2.38, "learning_rate": 6.939387877575516e-06, "loss": 0.1081, "step": 7659 }, { "epoch": 2.38, "learning_rate": 6.938987797559512e-06, "loss": 0.0956, "step": 7660 }, { "epoch": 2.38, "learning_rate": 6.938587717543509e-06, "loss": 0.0968, "step": 7661 }, { "epoch": 2.38, "learning_rate": 6.938187637527506e-06, "loss": 0.0672, "step": 7662 }, { "epoch": 2.38, "learning_rate": 6.937787557511502e-06, "loss": 0.0983, "step": 7663 }, { "epoch": 2.38, "learning_rate": 6.9373874774955e-06, "loss": 0.0793, "step": 7664 }, { "epoch": 2.38, "learning_rate": 6.936987397479497e-06, "loss": 0.1072, "step": 7665 }, { "epoch": 2.38, "learning_rate": 6.936587317463493e-06, "loss": 0.024, "step": 7666 }, { "epoch": 2.38, "learning_rate": 6.93618723744749e-06, "loss": 0.1779, "step": 7667 }, { "epoch": 2.38, "learning_rate": 6.9357871574314875e-06, "loss": 0.0988, "step": 7668 }, { "epoch": 2.38, "learning_rate": 6.935387077415484e-06, "loss": 0.2094, "step": 7669 }, { "epoch": 2.38, "learning_rate": 6.934986997399481e-06, "loss": 0.0469, "step": 7670 }, { "epoch": 2.38, "learning_rate": 6.934586917383477e-06, "loss": 0.0747, "step": 7671 }, { "epoch": 2.38, "learning_rate": 6.934186837367473e-06, "loss": 0.0451, "step": 7672 }, { "epoch": 2.38, "learning_rate": 6.933786757351471e-06, "loss": 0.0715, "step": 7673 }, { "epoch": 2.38, "learning_rate": 6.933386677335468e-06, "loss": 0.048, "step": 7674 }, { "epoch": 2.38, "learning_rate": 6.932986597319464e-06, "loss": 0.1334, "step": 7675 }, { "epoch": 2.38, "learning_rate": 6.932586517303461e-06, "loss": 0.0483, "step": 7676 }, { "epoch": 2.38, "learning_rate": 6.932186437287458e-06, "loss": 0.036, "step": 7677 }, { "epoch": 2.38, "learning_rate": 6.931786357271455e-06, "loss": 0.0843, "step": 7678 }, { "epoch": 2.38, "learning_rate": 6.931386277255452e-06, "loss": 0.0955, "step": 7679 }, { "epoch": 2.38, "learning_rate": 6.930986197239449e-06, "loss": 0.2071, "step": 7680 }, { "epoch": 2.38, "learning_rate": 6.930586117223445e-06, "loss": 0.0618, "step": 7681 }, { "epoch": 2.38, "learning_rate": 6.930186037207443e-06, "loss": 0.1121, "step": 7682 }, { "epoch": 2.38, "learning_rate": 6.929785957191439e-06, "loss": 0.163, "step": 7683 }, { "epoch": 2.38, "learning_rate": 6.929385877175435e-06, "loss": 0.0438, "step": 7684 }, { "epoch": 2.39, "learning_rate": 6.928985797159432e-06, "loss": 0.0909, "step": 7685 }, { "epoch": 2.39, "learning_rate": 6.928585717143429e-06, "loss": 0.0588, "step": 7686 }, { "epoch": 2.39, "learning_rate": 6.928185637127426e-06, "loss": 0.1704, "step": 7687 }, { "epoch": 2.39, "learning_rate": 6.927785557111423e-06, "loss": 0.1171, "step": 7688 }, { "epoch": 2.39, "learning_rate": 6.92738547709542e-06, "loss": 0.1208, "step": 7689 }, { "epoch": 2.39, "learning_rate": 6.926985397079416e-06, "loss": 0.0692, "step": 7690 }, { "epoch": 2.39, "learning_rate": 6.926585317063413e-06, "loss": 0.0405, "step": 7691 }, { "epoch": 2.39, "learning_rate": 6.9261852370474105e-06, "loss": 0.0738, "step": 7692 }, { "epoch": 2.39, "learning_rate": 6.925785157031407e-06, "loss": 0.0551, "step": 7693 }, { "epoch": 2.39, "learning_rate": 6.925385077015404e-06, "loss": 0.0469, "step": 7694 }, { "epoch": 2.39, "learning_rate": 6.924984996999401e-06, "loss": 0.143, "step": 7695 }, { "epoch": 2.39, "learning_rate": 6.9245849169833964e-06, "loss": 0.0949, "step": 7696 }, { "epoch": 2.39, "learning_rate": 6.924184836967394e-06, "loss": 0.0718, "step": 7697 }, { "epoch": 2.39, "learning_rate": 6.923784756951391e-06, "loss": 0.065, "step": 7698 }, { "epoch": 2.39, "learning_rate": 6.923384676935387e-06, "loss": 0.0592, "step": 7699 }, { "epoch": 2.39, "learning_rate": 6.922984596919384e-06, "loss": 0.0805, "step": 7700 }, { "epoch": 2.39, "learning_rate": 6.9225845169033816e-06, "loss": 0.0773, "step": 7701 }, { "epoch": 2.39, "learning_rate": 6.922184436887378e-06, "loss": 0.0608, "step": 7702 }, { "epoch": 2.39, "learning_rate": 6.921784356871375e-06, "loss": 0.212, "step": 7703 }, { "epoch": 2.39, "learning_rate": 6.921384276855372e-06, "loss": 0.0902, "step": 7704 }, { "epoch": 2.39, "learning_rate": 6.920984196839369e-06, "loss": 0.0777, "step": 7705 }, { "epoch": 2.39, "learning_rate": 6.920584116823366e-06, "loss": 0.0232, "step": 7706 }, { "epoch": 2.39, "learning_rate": 6.9201840368073625e-06, "loss": 0.0736, "step": 7707 }, { "epoch": 2.39, "learning_rate": 6.919783956791359e-06, "loss": 0.0421, "step": 7708 }, { "epoch": 2.39, "learning_rate": 6.919383876775355e-06, "loss": 0.1443, "step": 7709 }, { "epoch": 2.39, "learning_rate": 6.918983796759352e-06, "loss": 0.1048, "step": 7710 }, { "epoch": 2.39, "learning_rate": 6.918583716743349e-06, "loss": 0.1388, "step": 7711 }, { "epoch": 2.39, "learning_rate": 6.918183636727346e-06, "loss": 0.0714, "step": 7712 }, { "epoch": 2.39, "learning_rate": 6.917783556711343e-06, "loss": 0.069, "step": 7713 }, { "epoch": 2.39, "learning_rate": 6.917383476695339e-06, "loss": 0.1195, "step": 7714 }, { "epoch": 2.39, "learning_rate": 6.916983396679337e-06, "loss": 0.046, "step": 7715 }, { "epoch": 2.39, "learning_rate": 6.9165833166633335e-06, "loss": 0.0596, "step": 7716 }, { "epoch": 2.4, "learning_rate": 6.91618323664733e-06, "loss": 0.0607, "step": 7717 }, { "epoch": 2.4, "learning_rate": 6.915783156631327e-06, "loss": 0.1331, "step": 7718 }, { "epoch": 2.4, "learning_rate": 6.9153830766153245e-06, "loss": 0.036, "step": 7719 }, { "epoch": 2.4, "learning_rate": 6.914982996599321e-06, "loss": 0.0751, "step": 7720 }, { "epoch": 2.4, "learning_rate": 6.914582916583317e-06, "loss": 0.0654, "step": 7721 }, { "epoch": 2.4, "learning_rate": 6.914182836567314e-06, "loss": 0.1199, "step": 7722 }, { "epoch": 2.4, "learning_rate": 6.91378275655131e-06, "loss": 0.0987, "step": 7723 }, { "epoch": 2.4, "learning_rate": 6.913382676535307e-06, "loss": 0.1688, "step": 7724 }, { "epoch": 2.4, "learning_rate": 6.9129825965193046e-06, "loss": 0.0939, "step": 7725 }, { "epoch": 2.4, "learning_rate": 6.912582516503301e-06, "loss": 0.0826, "step": 7726 }, { "epoch": 2.4, "learning_rate": 6.912182436487298e-06, "loss": 0.0972, "step": 7727 }, { "epoch": 2.4, "learning_rate": 6.911782356471295e-06, "loss": 0.0617, "step": 7728 }, { "epoch": 2.4, "learning_rate": 6.911382276455292e-06, "loss": 0.0788, "step": 7729 }, { "epoch": 2.4, "learning_rate": 6.910982196439289e-06, "loss": 0.1789, "step": 7730 }, { "epoch": 2.4, "learning_rate": 6.9105821164232855e-06, "loss": 0.0642, "step": 7731 }, { "epoch": 2.4, "learning_rate": 6.910182036407282e-06, "loss": 0.0587, "step": 7732 }, { "epoch": 2.4, "learning_rate": 6.90978195639128e-06, "loss": 0.0649, "step": 7733 }, { "epoch": 2.4, "learning_rate": 6.909381876375275e-06, "loss": 0.0874, "step": 7734 }, { "epoch": 2.4, "learning_rate": 6.908981796359272e-06, "loss": 0.0344, "step": 7735 }, { "epoch": 2.4, "learning_rate": 6.908581716343269e-06, "loss": 0.08, "step": 7736 }, { "epoch": 2.4, "learning_rate": 6.908181636327266e-06, "loss": 0.1027, "step": 7737 }, { "epoch": 2.4, "learning_rate": 6.907781556311262e-06, "loss": 0.0969, "step": 7738 }, { "epoch": 2.4, "learning_rate": 6.90738147629526e-06, "loss": 0.1336, "step": 7739 }, { "epoch": 2.4, "learning_rate": 6.9069813962792566e-06, "loss": 0.0642, "step": 7740 }, { "epoch": 2.4, "learning_rate": 6.906581316263253e-06, "loss": 0.0756, "step": 7741 }, { "epoch": 2.4, "learning_rate": 6.90618123624725e-06, "loss": 0.0727, "step": 7742 }, { "epoch": 2.4, "learning_rate": 6.9057811562312475e-06, "loss": 0.0965, "step": 7743 }, { "epoch": 2.4, "learning_rate": 6.905381076215244e-06, "loss": 0.1959, "step": 7744 }, { "epoch": 2.4, "learning_rate": 6.904980996199241e-06, "loss": 0.0742, "step": 7745 }, { "epoch": 2.4, "learning_rate": 6.904580916183237e-06, "loss": 0.0978, "step": 7746 }, { "epoch": 2.4, "learning_rate": 6.904180836167233e-06, "loss": 0.0677, "step": 7747 }, { "epoch": 2.4, "learning_rate": 6.90378075615123e-06, "loss": 0.0794, "step": 7748 }, { "epoch": 2.41, "learning_rate": 6.903380676135228e-06, "loss": 0.0675, "step": 7749 }, { "epoch": 2.41, "learning_rate": 6.902980596119224e-06, "loss": 0.0884, "step": 7750 }, { "epoch": 2.41, "learning_rate": 6.902580516103221e-06, "loss": 0.0622, "step": 7751 }, { "epoch": 2.41, "learning_rate": 6.902180436087218e-06, "loss": 0.191, "step": 7752 }, { "epoch": 2.41, "learning_rate": 6.901780356071215e-06, "loss": 0.1037, "step": 7753 }, { "epoch": 2.41, "learning_rate": 6.901380276055212e-06, "loss": 0.0964, "step": 7754 }, { "epoch": 2.41, "learning_rate": 6.9009801960392085e-06, "loss": 0.152, "step": 7755 }, { "epoch": 2.41, "learning_rate": 6.900580116023205e-06, "loss": 0.0633, "step": 7756 }, { "epoch": 2.41, "learning_rate": 6.900180036007203e-06, "loss": 0.2199, "step": 7757 }, { "epoch": 2.41, "learning_rate": 6.8997799559911995e-06, "loss": 0.049, "step": 7758 }, { "epoch": 2.41, "learning_rate": 6.899379875975195e-06, "loss": 0.0588, "step": 7759 }, { "epoch": 2.41, "learning_rate": 6.898979795959192e-06, "loss": 0.0561, "step": 7760 }, { "epoch": 2.41, "learning_rate": 6.898579715943189e-06, "loss": 0.2559, "step": 7761 }, { "epoch": 2.41, "learning_rate": 6.898179635927185e-06, "loss": 0.0332, "step": 7762 }, { "epoch": 2.41, "learning_rate": 6.897779555911183e-06, "loss": 0.1281, "step": 7763 }, { "epoch": 2.41, "learning_rate": 6.8973794758951796e-06, "loss": 0.0664, "step": 7764 }, { "epoch": 2.41, "learning_rate": 6.896979395879176e-06, "loss": 0.0886, "step": 7765 }, { "epoch": 2.41, "learning_rate": 6.896579315863173e-06, "loss": 0.1126, "step": 7766 }, { "epoch": 2.41, "learning_rate": 6.8961792358471705e-06, "loss": 0.0875, "step": 7767 }, { "epoch": 2.41, "learning_rate": 6.895779155831167e-06, "loss": 0.0375, "step": 7768 }, { "epoch": 2.41, "learning_rate": 6.895379075815164e-06, "loss": 0.1388, "step": 7769 }, { "epoch": 2.41, "learning_rate": 6.8949789957991605e-06, "loss": 0.0936, "step": 7770 }, { "epoch": 2.41, "learning_rate": 6.894578915783158e-06, "loss": 0.0567, "step": 7771 }, { "epoch": 2.41, "learning_rate": 6.894178835767154e-06, "loss": 0.0521, "step": 7772 }, { "epoch": 2.41, "learning_rate": 6.893778755751151e-06, "loss": 0.1116, "step": 7773 }, { "epoch": 2.41, "learning_rate": 6.893378675735147e-06, "loss": 0.0364, "step": 7774 }, { "epoch": 2.41, "learning_rate": 6.892978595719144e-06, "loss": 0.088, "step": 7775 }, { "epoch": 2.41, "learning_rate": 6.892578515703141e-06, "loss": 0.1143, "step": 7776 }, { "epoch": 2.41, "learning_rate": 6.892178435687138e-06, "loss": 0.0567, "step": 7777 }, { "epoch": 2.41, "learning_rate": 6.891778355671135e-06, "loss": 0.0297, "step": 7778 }, { "epoch": 2.41, "learning_rate": 6.8913782756551316e-06, "loss": 0.0689, "step": 7779 }, { "epoch": 2.41, "learning_rate": 6.890978195639128e-06, "loss": 0.0646, "step": 7780 }, { "epoch": 2.42, "learning_rate": 6.890578115623126e-06, "loss": 0.0719, "step": 7781 }, { "epoch": 2.42, "learning_rate": 6.8901780356071225e-06, "loss": 0.0945, "step": 7782 }, { "epoch": 2.42, "learning_rate": 6.889777955591119e-06, "loss": 0.0511, "step": 7783 }, { "epoch": 2.42, "learning_rate": 6.889377875575115e-06, "loss": 0.084, "step": 7784 }, { "epoch": 2.42, "learning_rate": 6.888977795559112e-06, "loss": 0.0425, "step": 7785 }, { "epoch": 2.42, "learning_rate": 6.888577715543109e-06, "loss": 0.0568, "step": 7786 }, { "epoch": 2.42, "learning_rate": 6.888177635527106e-06, "loss": 0.0876, "step": 7787 }, { "epoch": 2.42, "learning_rate": 6.887777555511103e-06, "loss": 0.0986, "step": 7788 }, { "epoch": 2.42, "learning_rate": 6.887377475495099e-06, "loss": 0.1268, "step": 7789 }, { "epoch": 2.42, "learning_rate": 6.886977395479097e-06, "loss": 0.0775, "step": 7790 }, { "epoch": 2.42, "learning_rate": 6.8865773154630935e-06, "loss": 0.0614, "step": 7791 }, { "epoch": 2.42, "learning_rate": 6.88617723544709e-06, "loss": 0.1009, "step": 7792 }, { "epoch": 2.42, "learning_rate": 6.885777155431087e-06, "loss": 0.0505, "step": 7793 }, { "epoch": 2.42, "learning_rate": 6.8853770754150835e-06, "loss": 0.0632, "step": 7794 }, { "epoch": 2.42, "learning_rate": 6.884976995399081e-06, "loss": 0.0576, "step": 7795 }, { "epoch": 2.42, "learning_rate": 6.884576915383078e-06, "loss": 0.0513, "step": 7796 }, { "epoch": 2.42, "learning_rate": 6.884176835367074e-06, "loss": 0.1016, "step": 7797 }, { "epoch": 2.42, "learning_rate": 6.88377675535107e-06, "loss": 0.1144, "step": 7798 }, { "epoch": 2.42, "learning_rate": 6.883376675335067e-06, "loss": 0.1665, "step": 7799 }, { "epoch": 2.42, "learning_rate": 6.8829765953190645e-06, "loss": 0.0943, "step": 7800 }, { "epoch": 2.42, "learning_rate": 6.882576515303061e-06, "loss": 0.1152, "step": 7801 }, { "epoch": 2.42, "learning_rate": 6.882176435287058e-06, "loss": 0.0673, "step": 7802 }, { "epoch": 2.42, "learning_rate": 6.8817763552710546e-06, "loss": 0.1031, "step": 7803 }, { "epoch": 2.42, "learning_rate": 6.881376275255052e-06, "loss": 0.0522, "step": 7804 }, { "epoch": 2.42, "learning_rate": 6.880976195239049e-06, "loss": 0.1486, "step": 7805 }, { "epoch": 2.42, "learning_rate": 6.8805761152230455e-06, "loss": 0.0875, "step": 7806 }, { "epoch": 2.42, "learning_rate": 6.880176035207042e-06, "loss": 0.0684, "step": 7807 }, { "epoch": 2.42, "learning_rate": 6.879775955191039e-06, "loss": 0.0575, "step": 7808 }, { "epoch": 2.42, "learning_rate": 6.879375875175035e-06, "loss": 0.0414, "step": 7809 }, { "epoch": 2.42, "learning_rate": 6.878975795159032e-06, "loss": 0.1459, "step": 7810 }, { "epoch": 2.42, "learning_rate": 6.878575715143029e-06, "loss": 0.1612, "step": 7811 }, { "epoch": 2.42, "learning_rate": 6.878175635127026e-06, "loss": 0.0714, "step": 7812 }, { "epoch": 2.42, "learning_rate": 6.877775555111022e-06, "loss": 0.0525, "step": 7813 }, { "epoch": 2.43, "learning_rate": 6.87737547509502e-06, "loss": 0.0706, "step": 7814 }, { "epoch": 2.43, "learning_rate": 6.8769753950790165e-06, "loss": 0.0561, "step": 7815 }, { "epoch": 2.43, "learning_rate": 6.876575315063013e-06, "loss": 0.0843, "step": 7816 }, { "epoch": 2.43, "learning_rate": 6.87617523504701e-06, "loss": 0.1905, "step": 7817 }, { "epoch": 2.43, "learning_rate": 6.875775155031007e-06, "loss": 0.0703, "step": 7818 }, { "epoch": 2.43, "learning_rate": 6.875375075015004e-06, "loss": 0.0725, "step": 7819 }, { "epoch": 2.43, "learning_rate": 6.874974994999001e-06, "loss": 0.0668, "step": 7820 }, { "epoch": 2.43, "learning_rate": 6.8745749149829975e-06, "loss": 0.0591, "step": 7821 }, { "epoch": 2.43, "learning_rate": 6.874174834966993e-06, "loss": 0.0459, "step": 7822 }, { "epoch": 2.43, "learning_rate": 6.87377475495099e-06, "loss": 0.225, "step": 7823 }, { "epoch": 2.43, "learning_rate": 6.8733746749349875e-06, "loss": 0.0553, "step": 7824 }, { "epoch": 2.43, "learning_rate": 6.872974594918984e-06, "loss": 0.0689, "step": 7825 }, { "epoch": 2.43, "learning_rate": 6.872574514902981e-06, "loss": 0.2278, "step": 7826 }, { "epoch": 2.43, "learning_rate": 6.872174434886978e-06, "loss": 0.052, "step": 7827 }, { "epoch": 2.43, "learning_rate": 6.871774354870975e-06, "loss": 0.1521, "step": 7828 }, { "epoch": 2.43, "learning_rate": 6.871374274854972e-06, "loss": 0.063, "step": 7829 }, { "epoch": 2.43, "learning_rate": 6.8709741948389685e-06, "loss": 0.0562, "step": 7830 }, { "epoch": 2.43, "learning_rate": 6.870574114822965e-06, "loss": 0.2595, "step": 7831 }, { "epoch": 2.43, "learning_rate": 6.870174034806963e-06, "loss": 0.0682, "step": 7832 }, { "epoch": 2.43, "learning_rate": 6.869773954790959e-06, "loss": 0.0889, "step": 7833 }, { "epoch": 2.43, "learning_rate": 6.869373874774956e-06, "loss": 0.0733, "step": 7834 }, { "epoch": 2.43, "learning_rate": 6.868973794758952e-06, "loss": 0.072, "step": 7835 }, { "epoch": 2.43, "learning_rate": 6.868573714742949e-06, "loss": 0.092, "step": 7836 }, { "epoch": 2.43, "learning_rate": 6.868173634726945e-06, "loss": 0.2047, "step": 7837 }, { "epoch": 2.43, "learning_rate": 6.867773554710943e-06, "loss": 0.0392, "step": 7838 }, { "epoch": 2.43, "learning_rate": 6.8673734746949395e-06, "loss": 0.1003, "step": 7839 }, { "epoch": 2.43, "learning_rate": 6.866973394678936e-06, "loss": 0.1, "step": 7840 }, { "epoch": 2.43, "learning_rate": 6.866573314662933e-06, "loss": 0.0468, "step": 7841 }, { "epoch": 2.43, "learning_rate": 6.86617323464693e-06, "loss": 0.0786, "step": 7842 }, { "epoch": 2.43, "learning_rate": 6.865773154630927e-06, "loss": 0.0811, "step": 7843 }, { "epoch": 2.43, "learning_rate": 6.865373074614924e-06, "loss": 0.0907, "step": 7844 }, { "epoch": 2.43, "learning_rate": 6.8649729945989205e-06, "loss": 0.1664, "step": 7845 }, { "epoch": 2.44, "learning_rate": 6.864572914582918e-06, "loss": 0.0908, "step": 7846 }, { "epoch": 2.44, "learning_rate": 6.864172834566913e-06, "loss": 0.1544, "step": 7847 }, { "epoch": 2.44, "learning_rate": 6.8637727545509105e-06, "loss": 0.0805, "step": 7848 }, { "epoch": 2.44, "learning_rate": 6.863372674534907e-06, "loss": 0.1224, "step": 7849 }, { "epoch": 2.44, "learning_rate": 6.862972594518904e-06, "loss": 0.0401, "step": 7850 }, { "epoch": 2.44, "learning_rate": 6.862572514502901e-06, "loss": 0.0983, "step": 7851 }, { "epoch": 2.44, "learning_rate": 6.862172434486898e-06, "loss": 0.1019, "step": 7852 }, { "epoch": 2.44, "learning_rate": 6.861772354470895e-06, "loss": 0.0848, "step": 7853 }, { "epoch": 2.44, "learning_rate": 6.8613722744548915e-06, "loss": 0.1154, "step": 7854 }, { "epoch": 2.44, "learning_rate": 6.860972194438888e-06, "loss": 0.0957, "step": 7855 }, { "epoch": 2.44, "learning_rate": 6.860572114422886e-06, "loss": 0.1091, "step": 7856 }, { "epoch": 2.44, "learning_rate": 6.860172034406882e-06, "loss": 0.1229, "step": 7857 }, { "epoch": 2.44, "learning_rate": 6.859771954390879e-06, "loss": 0.0959, "step": 7858 }, { "epoch": 2.44, "learning_rate": 6.859371874374876e-06, "loss": 0.0458, "step": 7859 }, { "epoch": 2.44, "learning_rate": 6.858971794358872e-06, "loss": 0.0419, "step": 7860 }, { "epoch": 2.44, "learning_rate": 6.858571714342868e-06, "loss": 0.0717, "step": 7861 }, { "epoch": 2.44, "learning_rate": 6.858171634326866e-06, "loss": 0.1386, "step": 7862 }, { "epoch": 2.44, "learning_rate": 6.8577715543108625e-06, "loss": 0.108, "step": 7863 }, { "epoch": 2.44, "learning_rate": 6.857371474294859e-06, "loss": 0.0903, "step": 7864 }, { "epoch": 2.44, "learning_rate": 6.856971394278856e-06, "loss": 0.1636, "step": 7865 }, { "epoch": 2.44, "learning_rate": 6.8565713142628534e-06, "loss": 0.0919, "step": 7866 }, { "epoch": 2.44, "learning_rate": 6.85617123424685e-06, "loss": 0.0852, "step": 7867 }, { "epoch": 2.44, "learning_rate": 6.855771154230847e-06, "loss": 0.0674, "step": 7868 }, { "epoch": 2.44, "learning_rate": 6.8553710742148435e-06, "loss": 0.1199, "step": 7869 }, { "epoch": 2.44, "learning_rate": 6.854970994198841e-06, "loss": 0.2141, "step": 7870 }, { "epoch": 2.44, "learning_rate": 6.854570914182838e-06, "loss": 0.0371, "step": 7871 }, { "epoch": 2.44, "learning_rate": 6.8541708341668335e-06, "loss": 0.1106, "step": 7872 }, { "epoch": 2.44, "learning_rate": 6.85377075415083e-06, "loss": 0.1023, "step": 7873 }, { "epoch": 2.44, "learning_rate": 6.853370674134827e-06, "loss": 0.1306, "step": 7874 }, { "epoch": 2.44, "learning_rate": 6.8529705941188245e-06, "loss": 0.0764, "step": 7875 }, { "epoch": 2.44, "learning_rate": 6.852570514102821e-06, "loss": 0.0419, "step": 7876 }, { "epoch": 2.44, "learning_rate": 6.852170434086818e-06, "loss": 0.1268, "step": 7877 }, { "epoch": 2.45, "learning_rate": 6.8517703540708145e-06, "loss": 0.063, "step": 7878 }, { "epoch": 2.45, "learning_rate": 6.851370274054811e-06, "loss": 0.0618, "step": 7879 }, { "epoch": 2.45, "learning_rate": 6.850970194038809e-06, "loss": 0.0771, "step": 7880 }, { "epoch": 2.45, "learning_rate": 6.850570114022805e-06, "loss": 0.0418, "step": 7881 }, { "epoch": 2.45, "learning_rate": 6.850170034006802e-06, "loss": 0.0844, "step": 7882 }, { "epoch": 2.45, "learning_rate": 6.849769953990799e-06, "loss": 0.0484, "step": 7883 }, { "epoch": 2.45, "learning_rate": 6.849369873974796e-06, "loss": 0.044, "step": 7884 }, { "epoch": 2.45, "learning_rate": 6.848969793958792e-06, "loss": 0.0582, "step": 7885 }, { "epoch": 2.45, "learning_rate": 6.848569713942789e-06, "loss": 0.0895, "step": 7886 }, { "epoch": 2.45, "learning_rate": 6.8481696339267855e-06, "loss": 0.242, "step": 7887 }, { "epoch": 2.45, "learning_rate": 6.847769553910782e-06, "loss": 0.1412, "step": 7888 }, { "epoch": 2.45, "learning_rate": 6.84736947389478e-06, "loss": 0.0555, "step": 7889 }, { "epoch": 2.45, "learning_rate": 6.8469693938787764e-06, "loss": 0.2014, "step": 7890 }, { "epoch": 2.45, "learning_rate": 6.846569313862773e-06, "loss": 0.1427, "step": 7891 }, { "epoch": 2.45, "learning_rate": 6.84616923384677e-06, "loss": 0.0613, "step": 7892 }, { "epoch": 2.45, "learning_rate": 6.8457691538307665e-06, "loss": 0.1599, "step": 7893 }, { "epoch": 2.45, "learning_rate": 6.845369073814764e-06, "loss": 0.0615, "step": 7894 }, { "epoch": 2.45, "learning_rate": 6.844968993798761e-06, "loss": 0.2072, "step": 7895 }, { "epoch": 2.45, "learning_rate": 6.844568913782757e-06, "loss": 0.0538, "step": 7896 }, { "epoch": 2.45, "learning_rate": 6.844168833766754e-06, "loss": 0.0793, "step": 7897 }, { "epoch": 2.45, "learning_rate": 6.84376875375075e-06, "loss": 0.0803, "step": 7898 }, { "epoch": 2.45, "learning_rate": 6.8433686737347475e-06, "loss": 0.0284, "step": 7899 }, { "epoch": 2.45, "learning_rate": 6.842968593718744e-06, "loss": 0.0573, "step": 7900 }, { "epoch": 2.45, "learning_rate": 6.842568513702741e-06, "loss": 0.0454, "step": 7901 }, { "epoch": 2.45, "learning_rate": 6.8421684336867375e-06, "loss": 0.0579, "step": 7902 }, { "epoch": 2.45, "learning_rate": 6.841768353670735e-06, "loss": 0.0443, "step": 7903 }, { "epoch": 2.45, "learning_rate": 6.841368273654732e-06, "loss": 0.0861, "step": 7904 }, { "epoch": 2.45, "learning_rate": 6.8409681936387284e-06, "loss": 0.1023, "step": 7905 }, { "epoch": 2.45, "learning_rate": 6.840568113622725e-06, "loss": 0.0682, "step": 7906 }, { "epoch": 2.45, "learning_rate": 6.840168033606722e-06, "loss": 0.0625, "step": 7907 }, { "epoch": 2.45, "learning_rate": 6.839767953590719e-06, "loss": 0.1261, "step": 7908 }, { "epoch": 2.45, "learning_rate": 6.839367873574716e-06, "loss": 0.1115, "step": 7909 }, { "epoch": 2.46, "learning_rate": 6.838967793558712e-06, "loss": 0.0297, "step": 7910 }, { "epoch": 2.46, "learning_rate": 6.8385677135427085e-06, "loss": 0.0406, "step": 7911 }, { "epoch": 2.46, "learning_rate": 6.838167633526705e-06, "loss": 0.0906, "step": 7912 }, { "epoch": 2.46, "learning_rate": 6.837767553510703e-06, "loss": 0.2928, "step": 7913 }, { "epoch": 2.46, "learning_rate": 6.8373674734946995e-06, "loss": 0.112, "step": 7914 }, { "epoch": 2.46, "learning_rate": 6.836967393478696e-06, "loss": 0.0456, "step": 7915 }, { "epoch": 2.46, "learning_rate": 6.836567313462693e-06, "loss": 0.038, "step": 7916 }, { "epoch": 2.46, "learning_rate": 6.83616723344669e-06, "loss": 0.0706, "step": 7917 }, { "epoch": 2.46, "learning_rate": 6.835767153430687e-06, "loss": 0.0411, "step": 7918 }, { "epoch": 2.46, "learning_rate": 6.835367073414684e-06, "loss": 0.0846, "step": 7919 }, { "epoch": 2.46, "learning_rate": 6.83496699339868e-06, "loss": 0.0843, "step": 7920 }, { "epoch": 2.46, "learning_rate": 6.834566913382677e-06, "loss": 0.1531, "step": 7921 }, { "epoch": 2.46, "learning_rate": 6.834166833366675e-06, "loss": 0.0325, "step": 7922 }, { "epoch": 2.46, "learning_rate": 6.8337667533506705e-06, "loss": 0.0868, "step": 7923 }, { "epoch": 2.46, "learning_rate": 6.833366673334667e-06, "loss": 0.0649, "step": 7924 }, { "epoch": 2.46, "learning_rate": 6.832966593318664e-06, "loss": 0.236, "step": 7925 }, { "epoch": 2.46, "learning_rate": 6.8325665133026605e-06, "loss": 0.065, "step": 7926 }, { "epoch": 2.46, "learning_rate": 6.832166433286658e-06, "loss": 0.0563, "step": 7927 }, { "epoch": 2.46, "learning_rate": 6.831766353270655e-06, "loss": 0.0909, "step": 7928 }, { "epoch": 2.46, "learning_rate": 6.8313662732546514e-06, "loss": 0.1654, "step": 7929 }, { "epoch": 2.46, "learning_rate": 6.830966193238648e-06, "loss": 0.0399, "step": 7930 }, { "epoch": 2.46, "learning_rate": 6.830566113222646e-06, "loss": 0.1, "step": 7931 }, { "epoch": 2.46, "learning_rate": 6.830166033206642e-06, "loss": 0.0871, "step": 7932 }, { "epoch": 2.46, "learning_rate": 6.829765953190639e-06, "loss": 0.1141, "step": 7933 }, { "epoch": 2.46, "learning_rate": 6.829365873174636e-06, "loss": 0.0759, "step": 7934 }, { "epoch": 2.46, "learning_rate": 6.828965793158633e-06, "loss": 0.0423, "step": 7935 }, { "epoch": 2.46, "learning_rate": 6.828565713142628e-06, "loss": 0.0283, "step": 7936 }, { "epoch": 2.46, "learning_rate": 6.828165633126626e-06, "loss": 0.1009, "step": 7937 }, { "epoch": 2.46, "learning_rate": 6.8277655531106225e-06, "loss": 0.0886, "step": 7938 }, { "epoch": 2.46, "learning_rate": 6.827365473094619e-06, "loss": 0.0735, "step": 7939 }, { "epoch": 2.46, "learning_rate": 6.826965393078616e-06, "loss": 0.0601, "step": 7940 }, { "epoch": 2.46, "learning_rate": 6.826565313062613e-06, "loss": 0.061, "step": 7941 }, { "epoch": 2.47, "learning_rate": 6.82616523304661e-06, "loss": 0.0599, "step": 7942 }, { "epoch": 2.47, "learning_rate": 6.825765153030607e-06, "loss": 0.0681, "step": 7943 }, { "epoch": 2.47, "learning_rate": 6.8253650730146034e-06, "loss": 0.053, "step": 7944 }, { "epoch": 2.47, "learning_rate": 6.824964992998601e-06, "loss": 0.1749, "step": 7945 }, { "epoch": 2.47, "learning_rate": 6.824564912982598e-06, "loss": 0.0698, "step": 7946 }, { "epoch": 2.47, "learning_rate": 6.824164832966594e-06, "loss": 0.0558, "step": 7947 }, { "epoch": 2.47, "learning_rate": 6.82376475295059e-06, "loss": 0.1179, "step": 7948 }, { "epoch": 2.47, "learning_rate": 6.823364672934587e-06, "loss": 0.1397, "step": 7949 }, { "epoch": 2.47, "learning_rate": 6.8229645929185835e-06, "loss": 0.0839, "step": 7950 }, { "epoch": 2.47, "learning_rate": 6.822564512902581e-06, "loss": 0.0649, "step": 7951 }, { "epoch": 2.47, "learning_rate": 6.822164432886578e-06, "loss": 0.1, "step": 7952 }, { "epoch": 2.47, "learning_rate": 6.8217643528705745e-06, "loss": 0.0591, "step": 7953 }, { "epoch": 2.47, "learning_rate": 6.821364272854571e-06, "loss": 0.0735, "step": 7954 }, { "epoch": 2.47, "learning_rate": 6.820964192838569e-06, "loss": 0.0676, "step": 7955 }, { "epoch": 2.47, "learning_rate": 6.820564112822565e-06, "loss": 0.1824, "step": 7956 }, { "epoch": 2.47, "learning_rate": 6.820164032806562e-06, "loss": 0.0448, "step": 7957 }, { "epoch": 2.47, "learning_rate": 6.819763952790559e-06, "loss": 0.0743, "step": 7958 }, { "epoch": 2.47, "learning_rate": 6.819363872774556e-06, "loss": 0.0698, "step": 7959 }, { "epoch": 2.47, "learning_rate": 6.818963792758553e-06, "loss": 0.0449, "step": 7960 }, { "epoch": 2.47, "learning_rate": 6.818563712742549e-06, "loss": 0.0909, "step": 7961 }, { "epoch": 2.47, "learning_rate": 6.8181636327265455e-06, "loss": 0.0792, "step": 7962 }, { "epoch": 2.47, "learning_rate": 6.817763552710542e-06, "loss": 0.0693, "step": 7963 }, { "epoch": 2.47, "learning_rate": 6.817363472694539e-06, "loss": 0.0689, "step": 7964 }, { "epoch": 2.47, "learning_rate": 6.816963392678536e-06, "loss": 0.0867, "step": 7965 }, { "epoch": 2.47, "learning_rate": 6.816563312662533e-06, "loss": 0.1029, "step": 7966 }, { "epoch": 2.47, "learning_rate": 6.81616323264653e-06, "loss": 0.1323, "step": 7967 }, { "epoch": 2.47, "learning_rate": 6.8157631526305264e-06, "loss": 0.157, "step": 7968 }, { "epoch": 2.47, "learning_rate": 6.815363072614524e-06, "loss": 0.1714, "step": 7969 }, { "epoch": 2.47, "learning_rate": 6.814962992598521e-06, "loss": 0.0658, "step": 7970 }, { "epoch": 2.47, "learning_rate": 6.814562912582517e-06, "loss": 0.0746, "step": 7971 }, { "epoch": 2.47, "learning_rate": 6.814162832566514e-06, "loss": 0.1442, "step": 7972 }, { "epoch": 2.47, "learning_rate": 6.81376275255051e-06, "loss": 0.0894, "step": 7973 }, { "epoch": 2.47, "learning_rate": 6.813362672534507e-06, "loss": 0.1216, "step": 7974 }, { "epoch": 2.48, "learning_rate": 6.812962592518504e-06, "loss": 0.1282, "step": 7975 }, { "epoch": 2.48, "learning_rate": 6.812562512502501e-06, "loss": 0.1337, "step": 7976 }, { "epoch": 2.48, "learning_rate": 6.8121624324864975e-06, "loss": 0.1687, "step": 7977 }, { "epoch": 2.48, "learning_rate": 6.811762352470494e-06, "loss": 0.0616, "step": 7978 }, { "epoch": 2.48, "learning_rate": 6.811362272454492e-06, "loss": 0.0831, "step": 7979 }, { "epoch": 2.48, "learning_rate": 6.810962192438488e-06, "loss": 0.0822, "step": 7980 }, { "epoch": 2.48, "learning_rate": 6.810562112422485e-06, "loss": 0.1341, "step": 7981 }, { "epoch": 2.48, "learning_rate": 6.810162032406482e-06, "loss": 0.0733, "step": 7982 }, { "epoch": 2.48, "learning_rate": 6.809761952390479e-06, "loss": 0.0759, "step": 7983 }, { "epoch": 2.48, "learning_rate": 6.809361872374476e-06, "loss": 0.0819, "step": 7984 }, { "epoch": 2.48, "learning_rate": 6.808961792358473e-06, "loss": 0.0718, "step": 7985 }, { "epoch": 2.48, "learning_rate": 6.8085617123424685e-06, "loss": 0.0336, "step": 7986 }, { "epoch": 2.48, "learning_rate": 6.808161632326465e-06, "loss": 0.0785, "step": 7987 }, { "epoch": 2.48, "learning_rate": 6.807761552310463e-06, "loss": 0.0467, "step": 7988 }, { "epoch": 2.48, "learning_rate": 6.807361472294459e-06, "loss": 0.0986, "step": 7989 }, { "epoch": 2.48, "learning_rate": 6.806961392278456e-06, "loss": 0.1018, "step": 7990 }, { "epoch": 2.48, "learning_rate": 6.806561312262453e-06, "loss": 0.069, "step": 7991 }, { "epoch": 2.48, "learning_rate": 6.8061612322464495e-06, "loss": 0.066, "step": 7992 }, { "epoch": 2.48, "learning_rate": 6.805761152230447e-06, "loss": 0.0392, "step": 7993 }, { "epoch": 2.48, "learning_rate": 6.805361072214444e-06, "loss": 0.0347, "step": 7994 }, { "epoch": 2.48, "learning_rate": 6.80496099219844e-06, "loss": 0.0904, "step": 7995 }, { "epoch": 2.48, "learning_rate": 6.804560912182437e-06, "loss": 0.059, "step": 7996 }, { "epoch": 2.48, "learning_rate": 6.804160832166435e-06, "loss": 0.0506, "step": 7997 }, { "epoch": 2.48, "learning_rate": 6.803760752150431e-06, "loss": 0.0714, "step": 7998 }, { "epoch": 2.48, "learning_rate": 6.803360672134427e-06, "loss": 0.0748, "step": 7999 }, { "epoch": 2.48, "learning_rate": 6.802960592118424e-06, "loss": 0.0459, "step": 8000 }, { "epoch": 2.48, "learning_rate": 6.8025605121024205e-06, "loss": 0.0792, "step": 8001 }, { "epoch": 2.48, "learning_rate": 6.802160432086418e-06, "loss": 0.0632, "step": 8002 }, { "epoch": 2.48, "learning_rate": 6.801760352070415e-06, "loss": 0.1008, "step": 8003 }, { "epoch": 2.48, "learning_rate": 6.801360272054411e-06, "loss": 0.1372, "step": 8004 }, { "epoch": 2.48, "learning_rate": 6.800960192038408e-06, "loss": 0.0495, "step": 8005 }, { "epoch": 2.48, "learning_rate": 6.800560112022406e-06, "loss": 0.0876, "step": 8006 }, { "epoch": 2.49, "learning_rate": 6.800160032006402e-06, "loss": 0.081, "step": 8007 }, { "epoch": 2.49, "learning_rate": 6.799759951990399e-06, "loss": 0.0757, "step": 8008 }, { "epoch": 2.49, "learning_rate": 6.799359871974396e-06, "loss": 0.0721, "step": 8009 }, { "epoch": 2.49, "learning_rate": 6.798959791958392e-06, "loss": 0.0385, "step": 8010 }, { "epoch": 2.49, "learning_rate": 6.798559711942388e-06, "loss": 0.1305, "step": 8011 }, { "epoch": 2.49, "learning_rate": 6.798159631926386e-06, "loss": 0.0474, "step": 8012 }, { "epoch": 2.49, "learning_rate": 6.797759551910382e-06, "loss": 0.0487, "step": 8013 }, { "epoch": 2.49, "learning_rate": 6.797359471894379e-06, "loss": 0.085, "step": 8014 }, { "epoch": 2.49, "learning_rate": 6.796959391878376e-06, "loss": 0.0577, "step": 8015 }, { "epoch": 2.49, "learning_rate": 6.796559311862373e-06, "loss": 0.1076, "step": 8016 }, { "epoch": 2.49, "learning_rate": 6.79615923184637e-06, "loss": 0.1544, "step": 8017 }, { "epoch": 2.49, "learning_rate": 6.795759151830367e-06, "loss": 0.0302, "step": 8018 }, { "epoch": 2.49, "learning_rate": 6.795359071814363e-06, "loss": 0.0828, "step": 8019 }, { "epoch": 2.49, "learning_rate": 6.794958991798361e-06, "loss": 0.0416, "step": 8020 }, { "epoch": 2.49, "learning_rate": 6.794558911782358e-06, "loss": 0.0991, "step": 8021 }, { "epoch": 2.49, "learning_rate": 6.794158831766354e-06, "loss": 0.108, "step": 8022 }, { "epoch": 2.49, "learning_rate": 6.793758751750351e-06, "loss": 0.0892, "step": 8023 }, { "epoch": 2.49, "learning_rate": 6.793358671734347e-06, "loss": 0.0627, "step": 8024 }, { "epoch": 2.49, "learning_rate": 6.7929585917183435e-06, "loss": 0.2177, "step": 8025 }, { "epoch": 2.49, "learning_rate": 6.792558511702341e-06, "loss": 0.1072, "step": 8026 }, { "epoch": 2.49, "learning_rate": 6.792158431686338e-06, "loss": 0.044, "step": 8027 }, { "epoch": 2.49, "learning_rate": 6.791758351670334e-06, "loss": 0.0778, "step": 8028 }, { "epoch": 2.49, "learning_rate": 6.791358271654331e-06, "loss": 0.1547, "step": 8029 }, { "epoch": 2.49, "learning_rate": 6.790958191638329e-06, "loss": 0.0948, "step": 8030 }, { "epoch": 2.49, "learning_rate": 6.790558111622325e-06, "loss": 0.0757, "step": 8031 }, { "epoch": 2.49, "learning_rate": 6.790158031606322e-06, "loss": 0.055, "step": 8032 }, { "epoch": 2.49, "learning_rate": 6.789757951590319e-06, "loss": 0.02, "step": 8033 }, { "epoch": 2.49, "learning_rate": 6.789357871574316e-06, "loss": 0.0841, "step": 8034 }, { "epoch": 2.49, "learning_rate": 6.788957791558313e-06, "loss": 0.0849, "step": 8035 }, { "epoch": 2.49, "learning_rate": 6.788557711542309e-06, "loss": 0.0965, "step": 8036 }, { "epoch": 2.49, "learning_rate": 6.788157631526305e-06, "loss": 0.0589, "step": 8037 }, { "epoch": 2.49, "learning_rate": 6.787757551510302e-06, "loss": 0.0381, "step": 8038 }, { "epoch": 2.5, "learning_rate": 6.787357471494299e-06, "loss": 0.0698, "step": 8039 }, { "epoch": 2.5, "learning_rate": 6.786957391478296e-06, "loss": 0.1174, "step": 8040 }, { "epoch": 2.5, "learning_rate": 6.786557311462293e-06, "loss": 0.152, "step": 8041 }, { "epoch": 2.5, "learning_rate": 6.78615723144629e-06, "loss": 0.0875, "step": 8042 }, { "epoch": 2.5, "learning_rate": 6.785757151430286e-06, "loss": 0.0657, "step": 8043 }, { "epoch": 2.5, "learning_rate": 6.785357071414284e-06, "loss": 0.0679, "step": 8044 }, { "epoch": 2.5, "learning_rate": 6.784956991398281e-06, "loss": 0.0716, "step": 8045 }, { "epoch": 2.5, "learning_rate": 6.784556911382277e-06, "loss": 0.0218, "step": 8046 }, { "epoch": 2.5, "learning_rate": 6.784156831366274e-06, "loss": 0.087, "step": 8047 }, { "epoch": 2.5, "learning_rate": 6.7837567513502715e-06, "loss": 0.0579, "step": 8048 }, { "epoch": 2.5, "learning_rate": 6.7833566713342665e-06, "loss": 0.1012, "step": 8049 }, { "epoch": 2.5, "learning_rate": 6.782956591318264e-06, "loss": 0.2336, "step": 8050 }, { "epoch": 2.5, "learning_rate": 6.782556511302261e-06, "loss": 0.1008, "step": 8051 }, { "epoch": 2.5, "learning_rate": 6.782156431286257e-06, "loss": 0.1036, "step": 8052 }, { "epoch": 2.5, "learning_rate": 6.781756351270254e-06, "loss": 0.0895, "step": 8053 }, { "epoch": 2.5, "learning_rate": 6.781356271254252e-06, "loss": 0.1852, "step": 8054 }, { "epoch": 2.5, "learning_rate": 6.780956191238248e-06, "loss": 0.2741, "step": 8055 }, { "epoch": 2.5, "learning_rate": 6.780556111222245e-06, "loss": 0.0493, "step": 8056 }, { "epoch": 2.5, "learning_rate": 6.780156031206242e-06, "loss": 0.0789, "step": 8057 }, { "epoch": 2.5, "learning_rate": 6.779755951190239e-06, "loss": 0.182, "step": 8058 }, { "epoch": 2.5, "learning_rate": 6.779355871174236e-06, "loss": 0.0927, "step": 8059 }, { "epoch": 2.5, "learning_rate": 6.778955791158233e-06, "loss": 0.086, "step": 8060 }, { "epoch": 2.5, "learning_rate": 6.778555711142229e-06, "loss": 0.096, "step": 8061 }, { "epoch": 2.5, "learning_rate": 6.778155631126225e-06, "loss": 0.0746, "step": 8062 }, { "epoch": 2.5, "learning_rate": 6.777755551110222e-06, "loss": 0.035, "step": 8063 }, { "epoch": 2.5, "learning_rate": 6.777355471094219e-06, "loss": 0.0728, "step": 8064 }, { "epoch": 2.5, "learning_rate": 6.776955391078216e-06, "loss": 0.0781, "step": 8065 }, { "epoch": 2.5, "learning_rate": 6.776555311062213e-06, "loss": 0.0724, "step": 8066 }, { "epoch": 2.5, "learning_rate": 6.776155231046209e-06, "loss": 0.1673, "step": 8067 }, { "epoch": 2.5, "learning_rate": 6.775755151030207e-06, "loss": 0.0818, "step": 8068 }, { "epoch": 2.5, "learning_rate": 6.775355071014204e-06, "loss": 0.0587, "step": 8069 }, { "epoch": 2.5, "learning_rate": 6.7749549909982e-06, "loss": 0.0928, "step": 8070 }, { "epoch": 2.51, "learning_rate": 6.774554910982197e-06, "loss": 0.1189, "step": 8071 }, { "epoch": 2.51, "learning_rate": 6.7741548309661945e-06, "loss": 0.0959, "step": 8072 }, { "epoch": 2.51, "learning_rate": 6.773754750950191e-06, "loss": 0.1075, "step": 8073 }, { "epoch": 2.51, "learning_rate": 6.773354670934187e-06, "loss": 0.0796, "step": 8074 }, { "epoch": 2.51, "learning_rate": 6.772954590918184e-06, "loss": 0.0704, "step": 8075 }, { "epoch": 2.51, "learning_rate": 6.77255451090218e-06, "loss": 0.0814, "step": 8076 }, { "epoch": 2.51, "learning_rate": 6.772154430886177e-06, "loss": 0.0537, "step": 8077 }, { "epoch": 2.51, "learning_rate": 6.771754350870175e-06, "loss": 0.0372, "step": 8078 }, { "epoch": 2.51, "learning_rate": 6.771354270854171e-06, "loss": 0.0636, "step": 8079 }, { "epoch": 2.51, "learning_rate": 6.770954190838168e-06, "loss": 0.1421, "step": 8080 }, { "epoch": 2.51, "learning_rate": 6.770554110822165e-06, "loss": 0.1175, "step": 8081 }, { "epoch": 2.51, "learning_rate": 6.770154030806162e-06, "loss": 0.0482, "step": 8082 }, { "epoch": 2.51, "learning_rate": 6.769753950790159e-06, "loss": 0.0455, "step": 8083 }, { "epoch": 2.51, "learning_rate": 6.769353870774156e-06, "loss": 0.0633, "step": 8084 }, { "epoch": 2.51, "learning_rate": 6.768953790758152e-06, "loss": 0.0332, "step": 8085 }, { "epoch": 2.51, "learning_rate": 6.76855371074215e-06, "loss": 0.1422, "step": 8086 }, { "epoch": 2.51, "learning_rate": 6.768153630726146e-06, "loss": 0.0869, "step": 8087 }, { "epoch": 2.51, "learning_rate": 6.767753550710142e-06, "loss": 0.0604, "step": 8088 }, { "epoch": 2.51, "learning_rate": 6.767353470694139e-06, "loss": 0.0905, "step": 8089 }, { "epoch": 2.51, "learning_rate": 6.766953390678136e-06, "loss": 0.0514, "step": 8090 }, { "epoch": 2.51, "learning_rate": 6.766553310662133e-06, "loss": 0.0245, "step": 8091 }, { "epoch": 2.51, "learning_rate": 6.76615323064613e-06, "loss": 0.1499, "step": 8092 }, { "epoch": 2.51, "learning_rate": 6.765753150630127e-06, "loss": 0.0858, "step": 8093 }, { "epoch": 2.51, "learning_rate": 6.765353070614123e-06, "loss": 0.1112, "step": 8094 }, { "epoch": 2.51, "learning_rate": 6.76495299059812e-06, "loss": 0.1032, "step": 8095 }, { "epoch": 2.51, "learning_rate": 6.7645529105821175e-06, "loss": 0.0218, "step": 8096 }, { "epoch": 2.51, "learning_rate": 6.764152830566114e-06, "loss": 0.0497, "step": 8097 }, { "epoch": 2.51, "learning_rate": 6.763752750550111e-06, "loss": 0.0505, "step": 8098 }, { "epoch": 2.51, "learning_rate": 6.763352670534107e-06, "loss": 0.2245, "step": 8099 }, { "epoch": 2.51, "learning_rate": 6.7629525905181034e-06, "loss": 0.0953, "step": 8100 }, { "epoch": 2.51, "learning_rate": 6.762552510502101e-06, "loss": 0.1005, "step": 8101 }, { "epoch": 2.51, "learning_rate": 6.762152430486098e-06, "loss": 0.0557, "step": 8102 }, { "epoch": 2.51, "learning_rate": 6.761752350470094e-06, "loss": 0.0973, "step": 8103 }, { "epoch": 2.52, "learning_rate": 6.761352270454091e-06, "loss": 0.124, "step": 8104 }, { "epoch": 2.52, "learning_rate": 6.7609521904380886e-06, "loss": 0.0752, "step": 8105 }, { "epoch": 2.52, "learning_rate": 6.760552110422085e-06, "loss": 0.0562, "step": 8106 }, { "epoch": 2.52, "learning_rate": 6.760152030406082e-06, "loss": 0.1421, "step": 8107 }, { "epoch": 2.52, "learning_rate": 6.759751950390079e-06, "loss": 0.0684, "step": 8108 }, { "epoch": 2.52, "learning_rate": 6.759351870374075e-06, "loss": 0.0675, "step": 8109 }, { "epoch": 2.52, "learning_rate": 6.758951790358073e-06, "loss": 0.0505, "step": 8110 }, { "epoch": 2.52, "learning_rate": 6.7585517103420695e-06, "loss": 0.1003, "step": 8111 }, { "epoch": 2.52, "learning_rate": 6.758151630326065e-06, "loss": 0.0357, "step": 8112 }, { "epoch": 2.52, "learning_rate": 6.757751550310062e-06, "loss": 0.0909, "step": 8113 }, { "epoch": 2.52, "learning_rate": 6.757351470294059e-06, "loss": 0.0821, "step": 8114 }, { "epoch": 2.52, "learning_rate": 6.756951390278056e-06, "loss": 0.1554, "step": 8115 }, { "epoch": 2.52, "learning_rate": 6.756551310262053e-06, "loss": 0.1933, "step": 8116 }, { "epoch": 2.52, "learning_rate": 6.75615123024605e-06, "loss": 0.0978, "step": 8117 }, { "epoch": 2.52, "learning_rate": 6.755751150230046e-06, "loss": 0.0464, "step": 8118 }, { "epoch": 2.52, "learning_rate": 6.755351070214044e-06, "loss": 0.1278, "step": 8119 }, { "epoch": 2.52, "learning_rate": 6.7549509901980405e-06, "loss": 0.091, "step": 8120 }, { "epoch": 2.52, "learning_rate": 6.754550910182037e-06, "loss": 0.0635, "step": 8121 }, { "epoch": 2.52, "learning_rate": 6.754150830166034e-06, "loss": 0.0969, "step": 8122 }, { "epoch": 2.52, "learning_rate": 6.753750750150031e-06, "loss": 0.1419, "step": 8123 }, { "epoch": 2.52, "learning_rate": 6.753350670134028e-06, "loss": 0.078, "step": 8124 }, { "epoch": 2.52, "learning_rate": 6.752950590118024e-06, "loss": 0.0998, "step": 8125 }, { "epoch": 2.52, "learning_rate": 6.752550510102021e-06, "loss": 0.0726, "step": 8126 }, { "epoch": 2.52, "learning_rate": 6.752150430086017e-06, "loss": 0.1008, "step": 8127 }, { "epoch": 2.52, "learning_rate": 6.751750350070014e-06, "loss": 0.0738, "step": 8128 }, { "epoch": 2.52, "learning_rate": 6.7513502700540116e-06, "loss": 0.0864, "step": 8129 }, { "epoch": 2.52, "learning_rate": 6.750950190038008e-06, "loss": 0.1044, "step": 8130 }, { "epoch": 2.52, "learning_rate": 6.750550110022005e-06, "loss": 0.0749, "step": 8131 }, { "epoch": 2.52, "learning_rate": 6.750150030006002e-06, "loss": 0.0361, "step": 8132 }, { "epoch": 2.52, "learning_rate": 6.749749949989999e-06, "loss": 0.0754, "step": 8133 }, { "epoch": 2.52, "learning_rate": 6.749349869973996e-06, "loss": 0.0793, "step": 8134 }, { "epoch": 2.52, "learning_rate": 6.7489497899579925e-06, "loss": 0.0783, "step": 8135 }, { "epoch": 2.53, "learning_rate": 6.748549709941989e-06, "loss": 0.0465, "step": 8136 }, { "epoch": 2.53, "learning_rate": 6.748149629925985e-06, "loss": 0.1172, "step": 8137 }, { "epoch": 2.53, "learning_rate": 6.747749549909982e-06, "loss": 0.1793, "step": 8138 }, { "epoch": 2.53, "learning_rate": 6.747349469893979e-06, "loss": 0.0934, "step": 8139 }, { "epoch": 2.53, "learning_rate": 6.746949389877976e-06, "loss": 0.1037, "step": 8140 }, { "epoch": 2.53, "learning_rate": 6.746549309861973e-06, "loss": 0.1037, "step": 8141 }, { "epoch": 2.53, "learning_rate": 6.746149229845969e-06, "loss": 0.1694, "step": 8142 }, { "epoch": 2.53, "learning_rate": 6.745749149829967e-06, "loss": 0.1221, "step": 8143 }, { "epoch": 2.53, "learning_rate": 6.7453490698139636e-06, "loss": 0.0493, "step": 8144 }, { "epoch": 2.53, "learning_rate": 6.74494898979796e-06, "loss": 0.047, "step": 8145 }, { "epoch": 2.53, "learning_rate": 6.744548909781957e-06, "loss": 0.0508, "step": 8146 }, { "epoch": 2.53, "learning_rate": 6.7441488297659545e-06, "loss": 0.0553, "step": 8147 }, { "epoch": 2.53, "learning_rate": 6.743748749749951e-06, "loss": 0.0777, "step": 8148 }, { "epoch": 2.53, "learning_rate": 6.743348669733948e-06, "loss": 0.0538, "step": 8149 }, { "epoch": 2.53, "learning_rate": 6.742948589717944e-06, "loss": 0.0886, "step": 8150 }, { "epoch": 2.53, "learning_rate": 6.74254850970194e-06, "loss": 0.1285, "step": 8151 }, { "epoch": 2.53, "learning_rate": 6.742148429685937e-06, "loss": 0.0457, "step": 8152 }, { "epoch": 2.53, "learning_rate": 6.741748349669935e-06, "loss": 0.1149, "step": 8153 }, { "epoch": 2.53, "learning_rate": 6.741348269653931e-06, "loss": 0.0468, "step": 8154 }, { "epoch": 2.53, "learning_rate": 6.740948189637928e-06, "loss": 0.0646, "step": 8155 }, { "epoch": 2.53, "learning_rate": 6.740548109621925e-06, "loss": 0.1309, "step": 8156 }, { "epoch": 2.53, "learning_rate": 6.740148029605922e-06, "loss": 0.1016, "step": 8157 }, { "epoch": 2.53, "learning_rate": 6.739747949589919e-06, "loss": 0.0494, "step": 8158 }, { "epoch": 2.53, "learning_rate": 6.7393478695739155e-06, "loss": 0.0432, "step": 8159 }, { "epoch": 2.53, "learning_rate": 6.738947789557912e-06, "loss": 0.0509, "step": 8160 }, { "epoch": 2.53, "learning_rate": 6.73854770954191e-06, "loss": 0.1156, "step": 8161 }, { "epoch": 2.53, "learning_rate": 6.7381476295259064e-06, "loss": 0.2204, "step": 8162 }, { "epoch": 2.53, "learning_rate": 6.737747549509902e-06, "loss": 0.0997, "step": 8163 }, { "epoch": 2.53, "learning_rate": 6.737347469493899e-06, "loss": 0.0538, "step": 8164 }, { "epoch": 2.53, "learning_rate": 6.736947389477896e-06, "loss": 0.0878, "step": 8165 }, { "epoch": 2.53, "learning_rate": 6.736547309461892e-06, "loss": 0.1032, "step": 8166 }, { "epoch": 2.53, "learning_rate": 6.73614722944589e-06, "loss": 0.101, "step": 8167 }, { "epoch": 2.54, "learning_rate": 6.7357471494298866e-06, "loss": 0.0578, "step": 8168 }, { "epoch": 2.54, "learning_rate": 6.735347069413883e-06, "loss": 0.1087, "step": 8169 }, { "epoch": 2.54, "learning_rate": 6.73494698939788e-06, "loss": 0.1334, "step": 8170 }, { "epoch": 2.54, "learning_rate": 6.7345469093818775e-06, "loss": 0.0944, "step": 8171 }, { "epoch": 2.54, "learning_rate": 6.734146829365874e-06, "loss": 0.0774, "step": 8172 }, { "epoch": 2.54, "learning_rate": 6.733746749349871e-06, "loss": 0.2137, "step": 8173 }, { "epoch": 2.54, "learning_rate": 6.7333466693338675e-06, "loss": 0.1896, "step": 8174 }, { "epoch": 2.54, "learning_rate": 6.732946589317863e-06, "loss": 0.0576, "step": 8175 }, { "epoch": 2.54, "learning_rate": 6.732546509301861e-06, "loss": 0.0385, "step": 8176 }, { "epoch": 2.54, "learning_rate": 6.732146429285858e-06, "loss": 0.0578, "step": 8177 }, { "epoch": 2.54, "learning_rate": 6.731746349269854e-06, "loss": 0.091, "step": 8178 }, { "epoch": 2.54, "learning_rate": 6.731346269253851e-06, "loss": 0.1146, "step": 8179 }, { "epoch": 2.54, "learning_rate": 6.730946189237848e-06, "loss": 0.1032, "step": 8180 }, { "epoch": 2.54, "learning_rate": 6.730546109221845e-06, "loss": 0.0416, "step": 8181 }, { "epoch": 2.54, "learning_rate": 6.730146029205842e-06, "loss": 0.2465, "step": 8182 }, { "epoch": 2.54, "learning_rate": 6.7297459491898386e-06, "loss": 0.114, "step": 8183 }, { "epoch": 2.54, "learning_rate": 6.729345869173835e-06, "loss": 0.0806, "step": 8184 }, { "epoch": 2.54, "learning_rate": 6.728945789157833e-06, "loss": 0.0845, "step": 8185 }, { "epoch": 2.54, "learning_rate": 6.7285457091418295e-06, "loss": 0.0613, "step": 8186 }, { "epoch": 2.54, "learning_rate": 6.728145629125826e-06, "loss": 0.1141, "step": 8187 }, { "epoch": 2.54, "learning_rate": 6.727745549109822e-06, "loss": 0.0738, "step": 8188 }, { "epoch": 2.54, "learning_rate": 6.727345469093819e-06, "loss": 0.1085, "step": 8189 }, { "epoch": 2.54, "learning_rate": 6.726945389077816e-06, "loss": 0.032, "step": 8190 }, { "epoch": 2.54, "learning_rate": 6.726545309061813e-06, "loss": 0.1275, "step": 8191 }, { "epoch": 2.54, "learning_rate": 6.72614522904581e-06, "loss": 0.0761, "step": 8192 }, { "epoch": 2.54, "learning_rate": 6.725745149029806e-06, "loss": 0.1037, "step": 8193 }, { "epoch": 2.54, "learning_rate": 6.725345069013803e-06, "loss": 0.0512, "step": 8194 }, { "epoch": 2.54, "learning_rate": 6.7249449889978005e-06, "loss": 0.112, "step": 8195 }, { "epoch": 2.54, "learning_rate": 6.724544908981797e-06, "loss": 0.0375, "step": 8196 }, { "epoch": 2.54, "learning_rate": 6.724144828965794e-06, "loss": 0.0936, "step": 8197 }, { "epoch": 2.54, "learning_rate": 6.7237447489497905e-06, "loss": 0.087, "step": 8198 }, { "epoch": 2.54, "learning_rate": 6.723344668933788e-06, "loss": 0.1052, "step": 8199 }, { "epoch": 2.55, "learning_rate": 6.722944588917784e-06, "loss": 0.0696, "step": 8200 }, { "epoch": 2.55, "learning_rate": 6.722544508901781e-06, "loss": 0.0661, "step": 8201 }, { "epoch": 2.55, "learning_rate": 6.722144428885777e-06, "loss": 0.1187, "step": 8202 }, { "epoch": 2.55, "learning_rate": 6.721744348869774e-06, "loss": 0.041, "step": 8203 }, { "epoch": 2.55, "learning_rate": 6.7213442688537715e-06, "loss": 0.1701, "step": 8204 }, { "epoch": 2.55, "learning_rate": 6.720944188837768e-06, "loss": 0.0537, "step": 8205 }, { "epoch": 2.55, "learning_rate": 6.720544108821765e-06, "loss": 0.0772, "step": 8206 }, { "epoch": 2.55, "learning_rate": 6.7201440288057616e-06, "loss": 0.0885, "step": 8207 }, { "epoch": 2.55, "learning_rate": 6.719743948789758e-06, "loss": 0.1294, "step": 8208 }, { "epoch": 2.55, "learning_rate": 6.719343868773756e-06, "loss": 0.0593, "step": 8209 }, { "epoch": 2.55, "learning_rate": 6.7189437887577525e-06, "loss": 0.0847, "step": 8210 }, { "epoch": 2.55, "learning_rate": 6.718543708741749e-06, "loss": 0.1236, "step": 8211 }, { "epoch": 2.55, "learning_rate": 6.718143628725746e-06, "loss": 0.1284, "step": 8212 }, { "epoch": 2.55, "learning_rate": 6.717743548709742e-06, "loss": 0.1039, "step": 8213 }, { "epoch": 2.55, "learning_rate": 6.717343468693739e-06, "loss": 0.2103, "step": 8214 }, { "epoch": 2.55, "learning_rate": 6.716943388677736e-06, "loss": 0.0294, "step": 8215 }, { "epoch": 2.55, "learning_rate": 6.716543308661733e-06, "loss": 0.0985, "step": 8216 }, { "epoch": 2.55, "learning_rate": 6.716143228645729e-06, "loss": 0.0625, "step": 8217 }, { "epoch": 2.55, "learning_rate": 6.715743148629727e-06, "loss": 0.1156, "step": 8218 }, { "epoch": 2.55, "learning_rate": 6.7153430686137235e-06, "loss": 0.0549, "step": 8219 }, { "epoch": 2.55, "learning_rate": 6.71494298859772e-06, "loss": 0.1463, "step": 8220 }, { "epoch": 2.55, "learning_rate": 6.714542908581717e-06, "loss": 0.0915, "step": 8221 }, { "epoch": 2.55, "learning_rate": 6.7141428285657136e-06, "loss": 0.1066, "step": 8222 }, { "epoch": 2.55, "learning_rate": 6.713742748549711e-06, "loss": 0.0911, "step": 8223 }, { "epoch": 2.55, "learning_rate": 6.713342668533708e-06, "loss": 0.0512, "step": 8224 }, { "epoch": 2.55, "learning_rate": 6.7129425885177045e-06, "loss": 0.0504, "step": 8225 }, { "epoch": 2.55, "learning_rate": 6.7125425085017e-06, "loss": 0.0503, "step": 8226 }, { "epoch": 2.55, "learning_rate": 6.712142428485697e-06, "loss": 0.0515, "step": 8227 }, { "epoch": 2.55, "learning_rate": 6.7117423484696945e-06, "loss": 0.0494, "step": 8228 }, { "epoch": 2.55, "learning_rate": 6.711342268453691e-06, "loss": 0.0915, "step": 8229 }, { "epoch": 2.55, "learning_rate": 6.710942188437688e-06, "loss": 0.052, "step": 8230 }, { "epoch": 2.55, "learning_rate": 6.710542108421685e-06, "loss": 0.0781, "step": 8231 }, { "epoch": 2.56, "learning_rate": 6.710142028405682e-06, "loss": 0.0679, "step": 8232 }, { "epoch": 2.56, "learning_rate": 6.709741948389679e-06, "loss": 0.0436, "step": 8233 }, { "epoch": 2.56, "learning_rate": 6.7093418683736755e-06, "loss": 0.052, "step": 8234 }, { "epoch": 2.56, "learning_rate": 6.708941788357672e-06, "loss": 0.0786, "step": 8235 }, { "epoch": 2.56, "learning_rate": 6.70854170834167e-06, "loss": 0.1024, "step": 8236 }, { "epoch": 2.56, "learning_rate": 6.708141628325666e-06, "loss": 0.074, "step": 8237 }, { "epoch": 2.56, "learning_rate": 6.707741548309662e-06, "loss": 0.0626, "step": 8238 }, { "epoch": 2.56, "learning_rate": 6.707341468293659e-06, "loss": 0.0467, "step": 8239 }, { "epoch": 2.56, "learning_rate": 6.706941388277656e-06, "loss": 0.0628, "step": 8240 }, { "epoch": 2.56, "learning_rate": 6.706541308261652e-06, "loss": 0.1172, "step": 8241 }, { "epoch": 2.56, "learning_rate": 6.70614122824565e-06, "loss": 0.0813, "step": 8242 }, { "epoch": 2.56, "learning_rate": 6.7057411482296465e-06, "loss": 0.0716, "step": 8243 }, { "epoch": 2.56, "learning_rate": 6.705341068213643e-06, "loss": 0.0481, "step": 8244 }, { "epoch": 2.56, "learning_rate": 6.70494098819764e-06, "loss": 0.0754, "step": 8245 }, { "epoch": 2.56, "learning_rate": 6.704540908181637e-06, "loss": 0.038, "step": 8246 }, { "epoch": 2.56, "learning_rate": 6.704140828165634e-06, "loss": 0.1184, "step": 8247 }, { "epoch": 2.56, "learning_rate": 6.703740748149631e-06, "loss": 0.1026, "step": 8248 }, { "epoch": 2.56, "learning_rate": 6.7033406681336275e-06, "loss": 0.0839, "step": 8249 }, { "epoch": 2.56, "learning_rate": 6.702940588117625e-06, "loss": 0.0483, "step": 8250 }, { "epoch": 2.56, "learning_rate": 6.70254050810162e-06, "loss": 0.0633, "step": 8251 }, { "epoch": 2.56, "learning_rate": 6.7021404280856175e-06, "loss": 0.1132, "step": 8252 }, { "epoch": 2.56, "learning_rate": 6.701740348069614e-06, "loss": 0.1364, "step": 8253 }, { "epoch": 2.56, "learning_rate": 6.701340268053611e-06, "loss": 0.0566, "step": 8254 }, { "epoch": 2.56, "learning_rate": 6.700940188037608e-06, "loss": 0.0638, "step": 8255 }, { "epoch": 2.56, "learning_rate": 6.700540108021605e-06, "loss": 0.0705, "step": 8256 }, { "epoch": 2.56, "learning_rate": 6.700140028005602e-06, "loss": 0.0664, "step": 8257 }, { "epoch": 2.56, "learning_rate": 6.6997399479895985e-06, "loss": 0.1066, "step": 8258 }, { "epoch": 2.56, "learning_rate": 6.699339867973595e-06, "loss": 0.0444, "step": 8259 }, { "epoch": 2.56, "learning_rate": 6.698939787957593e-06, "loss": 0.0822, "step": 8260 }, { "epoch": 2.56, "learning_rate": 6.698539707941589e-06, "loss": 0.0658, "step": 8261 }, { "epoch": 2.56, "learning_rate": 6.698139627925586e-06, "loss": 0.0622, "step": 8262 }, { "epoch": 2.56, "learning_rate": 6.697739547909582e-06, "loss": 0.0688, "step": 8263 }, { "epoch": 2.56, "learning_rate": 6.697339467893579e-06, "loss": 0.0673, "step": 8264 }, { "epoch": 2.57, "learning_rate": 6.696939387877575e-06, "loss": 0.114, "step": 8265 }, { "epoch": 2.57, "learning_rate": 6.696539307861573e-06, "loss": 0.0836, "step": 8266 }, { "epoch": 2.57, "learning_rate": 6.6961392278455695e-06, "loss": 0.0829, "step": 8267 }, { "epoch": 2.57, "learning_rate": 6.695739147829566e-06, "loss": 0.0613, "step": 8268 }, { "epoch": 2.57, "learning_rate": 6.695339067813563e-06, "loss": 0.0597, "step": 8269 }, { "epoch": 2.57, "learning_rate": 6.6949389877975604e-06, "loss": 0.098, "step": 8270 }, { "epoch": 2.57, "learning_rate": 6.694538907781557e-06, "loss": 0.1237, "step": 8271 }, { "epoch": 2.57, "learning_rate": 6.694138827765554e-06, "loss": 0.0699, "step": 8272 }, { "epoch": 2.57, "learning_rate": 6.6937387477495505e-06, "loss": 0.06, "step": 8273 }, { "epoch": 2.57, "learning_rate": 6.693338667733548e-06, "loss": 0.1559, "step": 8274 }, { "epoch": 2.57, "learning_rate": 6.692938587717545e-06, "loss": 0.0829, "step": 8275 }, { "epoch": 2.57, "learning_rate": 6.6925385077015405e-06, "loss": 0.1494, "step": 8276 }, { "epoch": 2.57, "learning_rate": 6.692138427685537e-06, "loss": 0.0671, "step": 8277 }, { "epoch": 2.57, "learning_rate": 6.691738347669534e-06, "loss": 0.2413, "step": 8278 }, { "epoch": 2.57, "learning_rate": 6.691338267653531e-06, "loss": 0.0372, "step": 8279 }, { "epoch": 2.57, "learning_rate": 6.690938187637528e-06, "loss": 0.0714, "step": 8280 }, { "epoch": 2.57, "learning_rate": 6.690538107621525e-06, "loss": 0.101, "step": 8281 }, { "epoch": 2.57, "learning_rate": 6.6901380276055215e-06, "loss": 0.1219, "step": 8282 }, { "epoch": 2.57, "learning_rate": 6.689737947589518e-06, "loss": 0.0439, "step": 8283 }, { "epoch": 2.57, "learning_rate": 6.689337867573516e-06, "loss": 0.0621, "step": 8284 }, { "epoch": 2.57, "learning_rate": 6.688937787557512e-06, "loss": 0.0464, "step": 8285 }, { "epoch": 2.57, "learning_rate": 6.688537707541509e-06, "loss": 0.0484, "step": 8286 }, { "epoch": 2.57, "learning_rate": 6.688137627525506e-06, "loss": 0.1615, "step": 8287 }, { "epoch": 2.57, "learning_rate": 6.687737547509503e-06, "loss": 0.0547, "step": 8288 }, { "epoch": 2.57, "learning_rate": 6.687337467493499e-06, "loss": 0.0426, "step": 8289 }, { "epoch": 2.57, "learning_rate": 6.686937387477496e-06, "loss": 0.1399, "step": 8290 }, { "epoch": 2.57, "learning_rate": 6.6865373074614925e-06, "loss": 0.1414, "step": 8291 }, { "epoch": 2.57, "learning_rate": 6.686137227445489e-06, "loss": 0.0786, "step": 8292 }, { "epoch": 2.57, "learning_rate": 6.685737147429486e-06, "loss": 0.0447, "step": 8293 }, { "epoch": 2.57, "learning_rate": 6.6853370674134834e-06, "loss": 0.0624, "step": 8294 }, { "epoch": 2.57, "learning_rate": 6.68493698739748e-06, "loss": 0.1829, "step": 8295 }, { "epoch": 2.57, "learning_rate": 6.684536907381477e-06, "loss": 0.2145, "step": 8296 }, { "epoch": 2.58, "learning_rate": 6.6841368273654735e-06, "loss": 0.0683, "step": 8297 }, { "epoch": 2.58, "learning_rate": 6.683736747349471e-06, "loss": 0.0743, "step": 8298 }, { "epoch": 2.58, "learning_rate": 6.683336667333468e-06, "loss": 0.1187, "step": 8299 }, { "epoch": 2.58, "learning_rate": 6.682936587317464e-06, "loss": 0.1546, "step": 8300 }, { "epoch": 2.58, "learning_rate": 6.68253650730146e-06, "loss": 0.0305, "step": 8301 }, { "epoch": 2.58, "learning_rate": 6.682136427285457e-06, "loss": 0.0721, "step": 8302 }, { "epoch": 2.58, "learning_rate": 6.6817363472694545e-06, "loss": 0.0875, "step": 8303 }, { "epoch": 2.58, "learning_rate": 6.681336267253451e-06, "loss": 0.0963, "step": 8304 }, { "epoch": 2.58, "learning_rate": 6.680936187237448e-06, "loss": 0.0837, "step": 8305 }, { "epoch": 2.58, "learning_rate": 6.6805361072214445e-06, "loss": 0.0397, "step": 8306 }, { "epoch": 2.58, "learning_rate": 6.680136027205442e-06, "loss": 0.0755, "step": 8307 }, { "epoch": 2.58, "learning_rate": 6.679735947189439e-06, "loss": 0.0352, "step": 8308 }, { "epoch": 2.58, "learning_rate": 6.6793358671734354e-06, "loss": 0.0604, "step": 8309 }, { "epoch": 2.58, "learning_rate": 6.678935787157432e-06, "loss": 0.0613, "step": 8310 }, { "epoch": 2.58, "learning_rate": 6.678535707141429e-06, "loss": 0.1049, "step": 8311 }, { "epoch": 2.58, "learning_rate": 6.678135627125426e-06, "loss": 0.1009, "step": 8312 }, { "epoch": 2.58, "learning_rate": 6.677735547109423e-06, "loss": 0.0217, "step": 8313 }, { "epoch": 2.58, "learning_rate": 6.677335467093419e-06, "loss": 0.1106, "step": 8314 }, { "epoch": 2.58, "learning_rate": 6.6769353870774155e-06, "loss": 0.0603, "step": 8315 }, { "epoch": 2.58, "learning_rate": 6.676535307061412e-06, "loss": 0.0759, "step": 8316 }, { "epoch": 2.58, "learning_rate": 6.67613522704541e-06, "loss": 0.033, "step": 8317 }, { "epoch": 2.58, "learning_rate": 6.6757351470294064e-06, "loss": 0.0728, "step": 8318 }, { "epoch": 2.58, "learning_rate": 6.675335067013403e-06, "loss": 0.1639, "step": 8319 }, { "epoch": 2.58, "learning_rate": 6.6749349869974e-06, "loss": 0.2211, "step": 8320 }, { "epoch": 2.58, "learning_rate": 6.674534906981397e-06, "loss": 0.0983, "step": 8321 }, { "epoch": 2.58, "learning_rate": 6.674134826965394e-06, "loss": 0.1117, "step": 8322 }, { "epoch": 2.58, "learning_rate": 6.673734746949391e-06, "loss": 0.0721, "step": 8323 }, { "epoch": 2.58, "learning_rate": 6.673334666933387e-06, "loss": 0.0777, "step": 8324 }, { "epoch": 2.58, "learning_rate": 6.672934586917384e-06, "loss": 0.0601, "step": 8325 }, { "epoch": 2.58, "learning_rate": 6.67253450690138e-06, "loss": 0.1777, "step": 8326 }, { "epoch": 2.58, "learning_rate": 6.6721344268853775e-06, "loss": 0.0425, "step": 8327 }, { "epoch": 2.58, "learning_rate": 6.671734346869374e-06, "loss": 0.0285, "step": 8328 }, { "epoch": 2.59, "learning_rate": 6.671334266853371e-06, "loss": 0.1192, "step": 8329 }, { "epoch": 2.59, "learning_rate": 6.6709341868373675e-06, "loss": 0.0394, "step": 8330 }, { "epoch": 2.59, "learning_rate": 6.670534106821365e-06, "loss": 0.1733, "step": 8331 }, { "epoch": 2.59, "learning_rate": 6.670134026805362e-06, "loss": 0.1461, "step": 8332 }, { "epoch": 2.59, "learning_rate": 6.6697339467893584e-06, "loss": 0.0828, "step": 8333 }, { "epoch": 2.59, "learning_rate": 6.669333866773355e-06, "loss": 0.0415, "step": 8334 }, { "epoch": 2.59, "learning_rate": 6.668933786757353e-06, "loss": 0.0703, "step": 8335 }, { "epoch": 2.59, "learning_rate": 6.668533706741349e-06, "loss": 0.0669, "step": 8336 }, { "epoch": 2.59, "learning_rate": 6.668133626725346e-06, "loss": 0.1175, "step": 8337 }, { "epoch": 2.59, "learning_rate": 6.667733546709343e-06, "loss": 0.0781, "step": 8338 }, { "epoch": 2.59, "learning_rate": 6.6673334666933386e-06, "loss": 0.0648, "step": 8339 }, { "epoch": 2.59, "learning_rate": 6.666933386677335e-06, "loss": 0.1064, "step": 8340 }, { "epoch": 2.59, "learning_rate": 6.666533306661333e-06, "loss": 0.1151, "step": 8341 }, { "epoch": 2.59, "learning_rate": 6.6661332266453295e-06, "loss": 0.14, "step": 8342 }, { "epoch": 2.59, "learning_rate": 6.665733146629326e-06, "loss": 0.0319, "step": 8343 }, { "epoch": 2.59, "learning_rate": 6.665333066613323e-06, "loss": 0.0714, "step": 8344 }, { "epoch": 2.59, "learning_rate": 6.66493298659732e-06, "loss": 0.0581, "step": 8345 }, { "epoch": 2.59, "learning_rate": 6.664532906581317e-06, "loss": 0.1615, "step": 8346 }, { "epoch": 2.59, "learning_rate": 6.664132826565314e-06, "loss": 0.1369, "step": 8347 }, { "epoch": 2.59, "learning_rate": 6.6637327465493104e-06, "loss": 0.1031, "step": 8348 }, { "epoch": 2.59, "learning_rate": 6.663332666533308e-06, "loss": 0.0417, "step": 8349 }, { "epoch": 2.59, "learning_rate": 6.662932586517305e-06, "loss": 0.0808, "step": 8350 }, { "epoch": 2.59, "learning_rate": 6.662532506501301e-06, "loss": 0.1242, "step": 8351 }, { "epoch": 2.59, "learning_rate": 6.662132426485297e-06, "loss": 0.0845, "step": 8352 }, { "epoch": 2.59, "learning_rate": 6.661732346469294e-06, "loss": 0.0524, "step": 8353 }, { "epoch": 2.59, "learning_rate": 6.6613322664532905e-06, "loss": 0.0698, "step": 8354 }, { "epoch": 2.59, "learning_rate": 6.660932186437288e-06, "loss": 0.0802, "step": 8355 }, { "epoch": 2.59, "learning_rate": 6.660532106421285e-06, "loss": 0.0769, "step": 8356 }, { "epoch": 2.59, "learning_rate": 6.6601320264052814e-06, "loss": 0.0972, "step": 8357 }, { "epoch": 2.59, "learning_rate": 6.659731946389278e-06, "loss": 0.1319, "step": 8358 }, { "epoch": 2.59, "learning_rate": 6.659331866373276e-06, "loss": 0.0307, "step": 8359 }, { "epoch": 2.59, "learning_rate": 6.658931786357272e-06, "loss": 0.0811, "step": 8360 }, { "epoch": 2.6, "learning_rate": 6.658531706341269e-06, "loss": 0.1093, "step": 8361 }, { "epoch": 2.6, "learning_rate": 6.658131626325266e-06, "loss": 0.0691, "step": 8362 }, { "epoch": 2.6, "learning_rate": 6.657731546309263e-06, "loss": 0.0897, "step": 8363 }, { "epoch": 2.6, "learning_rate": 6.657331466293258e-06, "loss": 0.0607, "step": 8364 }, { "epoch": 2.6, "learning_rate": 6.656931386277256e-06, "loss": 0.075, "step": 8365 }, { "epoch": 2.6, "learning_rate": 6.6565313062612525e-06, "loss": 0.1381, "step": 8366 }, { "epoch": 2.6, "learning_rate": 6.656131226245249e-06, "loss": 0.0658, "step": 8367 }, { "epoch": 2.6, "learning_rate": 6.655731146229246e-06, "loss": 0.0747, "step": 8368 }, { "epoch": 2.6, "learning_rate": 6.655331066213243e-06, "loss": 0.0488, "step": 8369 }, { "epoch": 2.6, "learning_rate": 6.65493098619724e-06, "loss": 0.0624, "step": 8370 }, { "epoch": 2.6, "learning_rate": 6.654530906181237e-06, "loss": 0.0987, "step": 8371 }, { "epoch": 2.6, "learning_rate": 6.6541308261652334e-06, "loss": 0.2884, "step": 8372 }, { "epoch": 2.6, "learning_rate": 6.653730746149231e-06, "loss": 0.1446, "step": 8373 }, { "epoch": 2.6, "learning_rate": 6.653330666133228e-06, "loss": 0.0633, "step": 8374 }, { "epoch": 2.6, "learning_rate": 6.652930586117224e-06, "loss": 0.0535, "step": 8375 }, { "epoch": 2.6, "learning_rate": 6.652530506101221e-06, "loss": 0.0874, "step": 8376 }, { "epoch": 2.6, "learning_rate": 6.652130426085217e-06, "loss": 0.0603, "step": 8377 }, { "epoch": 2.6, "learning_rate": 6.6517303460692136e-06, "loss": 0.1072, "step": 8378 }, { "epoch": 2.6, "learning_rate": 6.651330266053211e-06, "loss": 0.0765, "step": 8379 }, { "epoch": 2.6, "learning_rate": 6.650930186037208e-06, "loss": 0.0781, "step": 8380 }, { "epoch": 2.6, "learning_rate": 6.6505301060212045e-06, "loss": 0.0481, "step": 8381 }, { "epoch": 2.6, "learning_rate": 6.650130026005201e-06, "loss": 0.2407, "step": 8382 }, { "epoch": 2.6, "learning_rate": 6.649729945989199e-06, "loss": 0.0655, "step": 8383 }, { "epoch": 2.6, "learning_rate": 6.649329865973195e-06, "loss": 0.1714, "step": 8384 }, { "epoch": 2.6, "learning_rate": 6.648929785957192e-06, "loss": 0.0943, "step": 8385 }, { "epoch": 2.6, "learning_rate": 6.648529705941189e-06, "loss": 0.0637, "step": 8386 }, { "epoch": 2.6, "learning_rate": 6.648129625925186e-06, "loss": 0.0645, "step": 8387 }, { "epoch": 2.6, "learning_rate": 6.647729545909183e-06, "loss": 0.2488, "step": 8388 }, { "epoch": 2.6, "learning_rate": 6.64732946589318e-06, "loss": 0.09, "step": 8389 }, { "epoch": 2.6, "learning_rate": 6.6469293858771755e-06, "loss": 0.0405, "step": 8390 }, { "epoch": 2.6, "learning_rate": 6.646529305861172e-06, "loss": 0.0697, "step": 8391 }, { "epoch": 2.6, "learning_rate": 6.64612922584517e-06, "loss": 0.0725, "step": 8392 }, { "epoch": 2.61, "learning_rate": 6.645729145829166e-06, "loss": 0.1151, "step": 8393 }, { "epoch": 2.61, "learning_rate": 6.645329065813163e-06, "loss": 0.0885, "step": 8394 }, { "epoch": 2.61, "learning_rate": 6.64492898579716e-06, "loss": 0.1328, "step": 8395 }, { "epoch": 2.61, "learning_rate": 6.6445289057811564e-06, "loss": 0.0545, "step": 8396 }, { "epoch": 2.61, "learning_rate": 6.644128825765154e-06, "loss": 0.0516, "step": 8397 }, { "epoch": 2.61, "learning_rate": 6.643728745749151e-06, "loss": 0.0925, "step": 8398 }, { "epoch": 2.61, "learning_rate": 6.643328665733147e-06, "loss": 0.1646, "step": 8399 }, { "epoch": 2.61, "learning_rate": 6.642928585717144e-06, "loss": 0.0724, "step": 8400 }, { "epoch": 2.61, "learning_rate": 6.6425285057011416e-06, "loss": 0.2106, "step": 8401 }, { "epoch": 2.61, "learning_rate": 6.642128425685137e-06, "loss": 0.0949, "step": 8402 }, { "epoch": 2.61, "learning_rate": 6.641728345669134e-06, "loss": 0.0728, "step": 8403 }, { "epoch": 2.61, "learning_rate": 6.641328265653131e-06, "loss": 0.1722, "step": 8404 }, { "epoch": 2.61, "learning_rate": 6.6409281856371275e-06, "loss": 0.02, "step": 8405 }, { "epoch": 2.61, "learning_rate": 6.640528105621125e-06, "loss": 0.0986, "step": 8406 }, { "epoch": 2.61, "learning_rate": 6.640128025605122e-06, "loss": 0.2514, "step": 8407 }, { "epoch": 2.61, "learning_rate": 6.639727945589118e-06, "loss": 0.1429, "step": 8408 }, { "epoch": 2.61, "learning_rate": 6.639327865573115e-06, "loss": 0.0901, "step": 8409 }, { "epoch": 2.61, "learning_rate": 6.638927785557112e-06, "loss": 0.1473, "step": 8410 }, { "epoch": 2.61, "learning_rate": 6.638527705541109e-06, "loss": 0.0333, "step": 8411 }, { "epoch": 2.61, "learning_rate": 6.638127625525106e-06, "loss": 0.0838, "step": 8412 }, { "epoch": 2.61, "learning_rate": 6.637727545509103e-06, "loss": 0.1037, "step": 8413 }, { "epoch": 2.61, "learning_rate": 6.637327465493099e-06, "loss": 0.0551, "step": 8414 }, { "epoch": 2.61, "learning_rate": 6.636927385477095e-06, "loss": 0.0792, "step": 8415 }, { "epoch": 2.61, "learning_rate": 6.636527305461093e-06, "loss": 0.044, "step": 8416 }, { "epoch": 2.61, "learning_rate": 6.636127225445089e-06, "loss": 0.0956, "step": 8417 }, { "epoch": 2.61, "learning_rate": 6.635727145429086e-06, "loss": 0.0615, "step": 8418 }, { "epoch": 2.61, "learning_rate": 6.635327065413083e-06, "loss": 0.1085, "step": 8419 }, { "epoch": 2.61, "learning_rate": 6.63492698539708e-06, "loss": 0.0613, "step": 8420 }, { "epoch": 2.61, "learning_rate": 6.634526905381077e-06, "loss": 0.1719, "step": 8421 }, { "epoch": 2.61, "learning_rate": 6.634126825365074e-06, "loss": 0.1189, "step": 8422 }, { "epoch": 2.61, "learning_rate": 6.63372674534907e-06, "loss": 0.0559, "step": 8423 }, { "epoch": 2.61, "learning_rate": 6.633326665333067e-06, "loss": 0.0515, "step": 8424 }, { "epoch": 2.61, "learning_rate": 6.632926585317065e-06, "loss": 0.0726, "step": 8425 }, { "epoch": 2.62, "learning_rate": 6.632526505301061e-06, "loss": 0.1267, "step": 8426 }, { "epoch": 2.62, "learning_rate": 6.632126425285057e-06, "loss": 0.1798, "step": 8427 }, { "epoch": 2.62, "learning_rate": 6.631726345269054e-06, "loss": 0.0512, "step": 8428 }, { "epoch": 2.62, "learning_rate": 6.6313262652530505e-06, "loss": 0.0381, "step": 8429 }, { "epoch": 2.62, "learning_rate": 6.630926185237048e-06, "loss": 0.0681, "step": 8430 }, { "epoch": 2.62, "learning_rate": 6.630526105221045e-06, "loss": 0.0518, "step": 8431 }, { "epoch": 2.62, "learning_rate": 6.630126025205041e-06, "loss": 0.0546, "step": 8432 }, { "epoch": 2.62, "learning_rate": 6.629725945189038e-06, "loss": 0.0682, "step": 8433 }, { "epoch": 2.62, "learning_rate": 6.629325865173036e-06, "loss": 0.0782, "step": 8434 }, { "epoch": 2.62, "learning_rate": 6.628925785157032e-06, "loss": 0.0949, "step": 8435 }, { "epoch": 2.62, "learning_rate": 6.628525705141029e-06, "loss": 0.0273, "step": 8436 }, { "epoch": 2.62, "learning_rate": 6.628125625125026e-06, "loss": 0.0758, "step": 8437 }, { "epoch": 2.62, "learning_rate": 6.627725545109022e-06, "loss": 0.0945, "step": 8438 }, { "epoch": 2.62, "learning_rate": 6.62732546509302e-06, "loss": 0.1697, "step": 8439 }, { "epoch": 2.62, "learning_rate": 6.626925385077016e-06, "loss": 0.034, "step": 8440 }, { "epoch": 2.62, "learning_rate": 6.626525305061012e-06, "loss": 0.0361, "step": 8441 }, { "epoch": 2.62, "learning_rate": 6.626125225045009e-06, "loss": 0.0771, "step": 8442 }, { "epoch": 2.62, "learning_rate": 6.625725145029006e-06, "loss": 0.0729, "step": 8443 }, { "epoch": 2.62, "learning_rate": 6.625325065013003e-06, "loss": 0.0633, "step": 8444 }, { "epoch": 2.62, "learning_rate": 6.624924984997e-06, "loss": 0.0423, "step": 8445 }, { "epoch": 2.62, "learning_rate": 6.624524904980997e-06, "loss": 0.0983, "step": 8446 }, { "epoch": 2.62, "learning_rate": 6.624124824964993e-06, "loss": 0.0528, "step": 8447 }, { "epoch": 2.62, "learning_rate": 6.623724744948991e-06, "loss": 0.0716, "step": 8448 }, { "epoch": 2.62, "learning_rate": 6.623324664932988e-06, "loss": 0.1032, "step": 8449 }, { "epoch": 2.62, "learning_rate": 6.622924584916984e-06, "loss": 0.0457, "step": 8450 }, { "epoch": 2.62, "learning_rate": 6.622524504900981e-06, "loss": 0.0371, "step": 8451 }, { "epoch": 2.62, "learning_rate": 6.6221244248849785e-06, "loss": 0.0577, "step": 8452 }, { "epoch": 2.62, "learning_rate": 6.6217243448689735e-06, "loss": 0.0943, "step": 8453 }, { "epoch": 2.62, "learning_rate": 6.621324264852971e-06, "loss": 0.1002, "step": 8454 }, { "epoch": 2.62, "learning_rate": 6.620924184836968e-06, "loss": 0.0808, "step": 8455 }, { "epoch": 2.62, "learning_rate": 6.620524104820964e-06, "loss": 0.1095, "step": 8456 }, { "epoch": 2.62, "learning_rate": 6.620124024804961e-06, "loss": 0.1643, "step": 8457 }, { "epoch": 2.63, "learning_rate": 6.619723944788959e-06, "loss": 0.0792, "step": 8458 }, { "epoch": 2.63, "learning_rate": 6.619323864772955e-06, "loss": 0.0851, "step": 8459 }, { "epoch": 2.63, "learning_rate": 6.618923784756952e-06, "loss": 0.0705, "step": 8460 }, { "epoch": 2.63, "learning_rate": 6.618523704740949e-06, "loss": 0.0566, "step": 8461 }, { "epoch": 2.63, "learning_rate": 6.618123624724946e-06, "loss": 0.092, "step": 8462 }, { "epoch": 2.63, "learning_rate": 6.617723544708943e-06, "loss": 0.0739, "step": 8463 }, { "epoch": 2.63, "learning_rate": 6.61732346469294e-06, "loss": 0.0599, "step": 8464 }, { "epoch": 2.63, "learning_rate": 6.6169233846769354e-06, "loss": 0.1026, "step": 8465 }, { "epoch": 2.63, "learning_rate": 6.616523304660932e-06, "loss": 0.161, "step": 8466 }, { "epoch": 2.63, "learning_rate": 6.616123224644929e-06, "loss": 0.0652, "step": 8467 }, { "epoch": 2.63, "learning_rate": 6.615723144628926e-06, "loss": 0.1686, "step": 8468 }, { "epoch": 2.63, "learning_rate": 6.615323064612923e-06, "loss": 0.082, "step": 8469 }, { "epoch": 2.63, "learning_rate": 6.61492298459692e-06, "loss": 0.0566, "step": 8470 }, { "epoch": 2.63, "learning_rate": 6.614522904580916e-06, "loss": 0.0928, "step": 8471 }, { "epoch": 2.63, "learning_rate": 6.614122824564914e-06, "loss": 0.0473, "step": 8472 }, { "epoch": 2.63, "learning_rate": 6.613722744548911e-06, "loss": 0.2036, "step": 8473 }, { "epoch": 2.63, "learning_rate": 6.613322664532907e-06, "loss": 0.0902, "step": 8474 }, { "epoch": 2.63, "learning_rate": 6.612922584516904e-06, "loss": 0.0739, "step": 8475 }, { "epoch": 2.63, "learning_rate": 6.6125225045009015e-06, "loss": 0.0917, "step": 8476 }, { "epoch": 2.63, "learning_rate": 6.612122424484898e-06, "loss": 0.0985, "step": 8477 }, { "epoch": 2.63, "learning_rate": 6.611722344468894e-06, "loss": 0.1224, "step": 8478 }, { "epoch": 2.63, "learning_rate": 6.611322264452891e-06, "loss": 0.1273, "step": 8479 }, { "epoch": 2.63, "learning_rate": 6.610922184436887e-06, "loss": 0.1037, "step": 8480 }, { "epoch": 2.63, "learning_rate": 6.610522104420884e-06, "loss": 0.0481, "step": 8481 }, { "epoch": 2.63, "learning_rate": 6.610122024404882e-06, "loss": 0.0498, "step": 8482 }, { "epoch": 2.63, "learning_rate": 6.609721944388878e-06, "loss": 0.0402, "step": 8483 }, { "epoch": 2.63, "learning_rate": 6.609321864372875e-06, "loss": 0.037, "step": 8484 }, { "epoch": 2.63, "learning_rate": 6.608921784356872e-06, "loss": 0.0551, "step": 8485 }, { "epoch": 2.63, "learning_rate": 6.608521704340869e-06, "loss": 0.037, "step": 8486 }, { "epoch": 2.63, "learning_rate": 6.608121624324866e-06, "loss": 0.1281, "step": 8487 }, { "epoch": 2.63, "learning_rate": 6.607721544308863e-06, "loss": 0.0473, "step": 8488 }, { "epoch": 2.63, "learning_rate": 6.607321464292859e-06, "loss": 0.1263, "step": 8489 }, { "epoch": 2.64, "learning_rate": 6.606921384276855e-06, "loss": 0.0709, "step": 8490 }, { "epoch": 2.64, "learning_rate": 6.606521304260853e-06, "loss": 0.0458, "step": 8491 }, { "epoch": 2.64, "learning_rate": 6.606121224244849e-06, "loss": 0.079, "step": 8492 }, { "epoch": 2.64, "learning_rate": 6.605721144228846e-06, "loss": 0.1027, "step": 8493 }, { "epoch": 2.64, "learning_rate": 6.605321064212843e-06, "loss": 0.1063, "step": 8494 }, { "epoch": 2.64, "learning_rate": 6.605321064212843e-06, "loss": 0.0741, "step": 8495 }, { "epoch": 2.64, "learning_rate": 6.604920984196839e-06, "loss": 0.0362, "step": 8496 }, { "epoch": 2.64, "learning_rate": 6.604520904180837e-06, "loss": 0.0552, "step": 8497 }, { "epoch": 2.64, "learning_rate": 6.604120824164834e-06, "loss": 0.1176, "step": 8498 }, { "epoch": 2.64, "learning_rate": 6.60372074414883e-06, "loss": 0.0385, "step": 8499 }, { "epoch": 2.64, "learning_rate": 6.603320664132827e-06, "loss": 0.1081, "step": 8500 }, { "epoch": 2.64, "learning_rate": 6.6029205841168245e-06, "loss": 0.0773, "step": 8501 }, { "epoch": 2.64, "learning_rate": 6.602520504100821e-06, "loss": 0.0972, "step": 8502 }, { "epoch": 2.64, "learning_rate": 6.602120424084818e-06, "loss": 0.1122, "step": 8503 }, { "epoch": 2.64, "learning_rate": 6.601720344068814e-06, "loss": 0.0726, "step": 8504 }, { "epoch": 2.64, "learning_rate": 6.6013202640528104e-06, "loss": 0.0394, "step": 8505 }, { "epoch": 2.64, "learning_rate": 6.600920184036808e-06, "loss": 0.1151, "step": 8506 }, { "epoch": 2.64, "learning_rate": 6.600520104020805e-06, "loss": 0.0735, "step": 8507 }, { "epoch": 2.64, "learning_rate": 6.600120024004801e-06, "loss": 0.0327, "step": 8508 }, { "epoch": 2.64, "learning_rate": 6.599719943988798e-06, "loss": 0.0777, "step": 8509 }, { "epoch": 2.64, "learning_rate": 6.599319863972795e-06, "loss": 0.0866, "step": 8510 }, { "epoch": 2.64, "learning_rate": 6.598919783956792e-06, "loss": 0.0535, "step": 8511 }, { "epoch": 2.64, "learning_rate": 6.598519703940789e-06, "loss": 0.0859, "step": 8512 }, { "epoch": 2.64, "learning_rate": 6.598119623924786e-06, "loss": 0.0595, "step": 8513 }, { "epoch": 2.64, "learning_rate": 6.597719543908782e-06, "loss": 0.0689, "step": 8514 }, { "epoch": 2.64, "learning_rate": 6.59731946389278e-06, "loss": 0.0749, "step": 8515 }, { "epoch": 2.64, "learning_rate": 6.5969193838767765e-06, "loss": 0.0941, "step": 8516 }, { "epoch": 2.64, "learning_rate": 6.596519303860772e-06, "loss": 0.0872, "step": 8517 }, { "epoch": 2.64, "learning_rate": 6.596119223844769e-06, "loss": 0.0858, "step": 8518 }, { "epoch": 2.64, "learning_rate": 6.595719143828766e-06, "loss": 0.0991, "step": 8519 }, { "epoch": 2.64, "learning_rate": 6.595319063812763e-06, "loss": 0.0992, "step": 8520 }, { "epoch": 2.64, "learning_rate": 6.59491898379676e-06, "loss": 0.0664, "step": 8521 }, { "epoch": 2.65, "learning_rate": 6.594518903780757e-06, "loss": 0.0695, "step": 8522 }, { "epoch": 2.65, "learning_rate": 6.594118823764753e-06, "loss": 0.0399, "step": 8523 }, { "epoch": 2.65, "learning_rate": 6.59371874374875e-06, "loss": 0.066, "step": 8524 }, { "epoch": 2.65, "learning_rate": 6.5933186637327475e-06, "loss": 0.1481, "step": 8525 }, { "epoch": 2.65, "learning_rate": 6.592918583716744e-06, "loss": 0.0886, "step": 8526 }, { "epoch": 2.65, "learning_rate": 6.592518503700741e-06, "loss": 0.0535, "step": 8527 }, { "epoch": 2.65, "learning_rate": 6.592118423684738e-06, "loss": 0.1078, "step": 8528 }, { "epoch": 2.65, "learning_rate": 6.5917183436687334e-06, "loss": 0.0805, "step": 8529 }, { "epoch": 2.65, "learning_rate": 6.591318263652731e-06, "loss": 0.0942, "step": 8530 }, { "epoch": 2.65, "learning_rate": 6.590918183636728e-06, "loss": 0.0962, "step": 8531 }, { "epoch": 2.65, "learning_rate": 6.590518103620724e-06, "loss": 0.0879, "step": 8532 }, { "epoch": 2.65, "learning_rate": 6.590118023604721e-06, "loss": 0.1471, "step": 8533 }, { "epoch": 2.65, "learning_rate": 6.5897179435887186e-06, "loss": 0.0731, "step": 8534 }, { "epoch": 2.65, "learning_rate": 6.589317863572715e-06, "loss": 0.0742, "step": 8535 }, { "epoch": 2.65, "learning_rate": 6.588917783556712e-06, "loss": 0.0492, "step": 8536 }, { "epoch": 2.65, "learning_rate": 6.588517703540709e-06, "loss": 0.1074, "step": 8537 }, { "epoch": 2.65, "learning_rate": 6.588117623524706e-06, "loss": 0.1158, "step": 8538 }, { "epoch": 2.65, "learning_rate": 6.587717543508703e-06, "loss": 0.1282, "step": 8539 }, { "epoch": 2.65, "learning_rate": 6.5873174634926995e-06, "loss": 0.0867, "step": 8540 }, { "epoch": 2.65, "learning_rate": 6.586917383476696e-06, "loss": 0.1104, "step": 8541 }, { "epoch": 2.65, "learning_rate": 6.586517303460692e-06, "loss": 0.0623, "step": 8542 }, { "epoch": 2.65, "learning_rate": 6.586117223444689e-06, "loss": 0.0808, "step": 8543 }, { "epoch": 2.65, "learning_rate": 6.585717143428686e-06, "loss": 0.0909, "step": 8544 }, { "epoch": 2.65, "learning_rate": 6.585317063412683e-06, "loss": 0.0644, "step": 8545 }, { "epoch": 2.65, "learning_rate": 6.58491698339668e-06, "loss": 0.0693, "step": 8546 }, { "epoch": 2.65, "learning_rate": 6.584516903380676e-06, "loss": 0.0357, "step": 8547 }, { "epoch": 2.65, "learning_rate": 6.584116823364674e-06, "loss": 0.056, "step": 8548 }, { "epoch": 2.65, "learning_rate": 6.5837167433486706e-06, "loss": 0.0493, "step": 8549 }, { "epoch": 2.65, "learning_rate": 6.583316663332667e-06, "loss": 0.1064, "step": 8550 }, { "epoch": 2.65, "learning_rate": 6.582916583316664e-06, "loss": 0.1952, "step": 8551 }, { "epoch": 2.65, "learning_rate": 6.5825165033006615e-06, "loss": 0.0916, "step": 8552 }, { "epoch": 2.65, "learning_rate": 6.582116423284658e-06, "loss": 0.113, "step": 8553 }, { "epoch": 2.65, "learning_rate": 6.581716343268654e-06, "loss": 0.0465, "step": 8554 }, { "epoch": 2.66, "learning_rate": 6.581316263252651e-06, "loss": 0.1006, "step": 8555 }, { "epoch": 2.66, "learning_rate": 6.580916183236647e-06, "loss": 0.0803, "step": 8556 }, { "epoch": 2.66, "learning_rate": 6.580516103220644e-06, "loss": 0.1222, "step": 8557 }, { "epoch": 2.66, "learning_rate": 6.5801160232046416e-06, "loss": 0.1059, "step": 8558 }, { "epoch": 2.66, "learning_rate": 6.579715943188638e-06, "loss": 0.1042, "step": 8559 }, { "epoch": 2.66, "learning_rate": 6.579315863172635e-06, "loss": 0.0801, "step": 8560 }, { "epoch": 2.66, "learning_rate": 6.578915783156632e-06, "loss": 0.0662, "step": 8561 }, { "epoch": 2.66, "learning_rate": 6.578515703140629e-06, "loss": 0.0689, "step": 8562 }, { "epoch": 2.66, "learning_rate": 6.578115623124626e-06, "loss": 0.1484, "step": 8563 }, { "epoch": 2.66, "learning_rate": 6.5777155431086225e-06, "loss": 0.1015, "step": 8564 }, { "epoch": 2.66, "learning_rate": 6.577315463092619e-06, "loss": 0.0565, "step": 8565 }, { "epoch": 2.66, "learning_rate": 6.576915383076617e-06, "loss": 0.1207, "step": 8566 }, { "epoch": 2.66, "learning_rate": 6.576515303060612e-06, "loss": 0.0742, "step": 8567 }, { "epoch": 2.66, "learning_rate": 6.576115223044609e-06, "loss": 0.1731, "step": 8568 }, { "epoch": 2.66, "learning_rate": 6.575715143028606e-06, "loss": 0.2215, "step": 8569 }, { "epoch": 2.66, "learning_rate": 6.575315063012603e-06, "loss": 0.0892, "step": 8570 }, { "epoch": 2.66, "learning_rate": 6.574914982996599e-06, "loss": 0.0511, "step": 8571 }, { "epoch": 2.66, "learning_rate": 6.574514902980597e-06, "loss": 0.0741, "step": 8572 }, { "epoch": 2.66, "learning_rate": 6.5741148229645936e-06, "loss": 0.1006, "step": 8573 }, { "epoch": 2.66, "learning_rate": 6.57371474294859e-06, "loss": 0.0249, "step": 8574 }, { "epoch": 2.66, "learning_rate": 6.573314662932587e-06, "loss": 0.1151, "step": 8575 }, { "epoch": 2.66, "learning_rate": 6.5729145829165845e-06, "loss": 0.1112, "step": 8576 }, { "epoch": 2.66, "learning_rate": 6.572514502900581e-06, "loss": 0.1004, "step": 8577 }, { "epoch": 2.66, "learning_rate": 6.572114422884578e-06, "loss": 0.0372, "step": 8578 }, { "epoch": 2.66, "learning_rate": 6.5717143428685745e-06, "loss": 0.0492, "step": 8579 }, { "epoch": 2.66, "learning_rate": 6.57131426285257e-06, "loss": 0.0525, "step": 8580 }, { "epoch": 2.66, "learning_rate": 6.570914182836567e-06, "loss": 0.0753, "step": 8581 }, { "epoch": 2.66, "learning_rate": 6.570514102820565e-06, "loss": 0.1189, "step": 8582 }, { "epoch": 2.66, "learning_rate": 6.570114022804561e-06, "loss": 0.1101, "step": 8583 }, { "epoch": 2.66, "learning_rate": 6.569713942788558e-06, "loss": 0.0197, "step": 8584 }, { "epoch": 2.66, "learning_rate": 6.569313862772555e-06, "loss": 0.0806, "step": 8585 }, { "epoch": 2.66, "learning_rate": 6.568913782756552e-06, "loss": 0.3339, "step": 8586 }, { "epoch": 2.67, "learning_rate": 6.568513702740549e-06, "loss": 0.0832, "step": 8587 }, { "epoch": 2.67, "learning_rate": 6.5681136227245456e-06, "loss": 0.1217, "step": 8588 }, { "epoch": 2.67, "learning_rate": 6.567713542708542e-06, "loss": 0.0657, "step": 8589 }, { "epoch": 2.67, "learning_rate": 6.56731346269254e-06, "loss": 0.0791, "step": 8590 }, { "epoch": 2.67, "learning_rate": 6.5669133826765365e-06, "loss": 0.1052, "step": 8591 }, { "epoch": 2.67, "learning_rate": 6.566513302660532e-06, "loss": 0.1078, "step": 8592 }, { "epoch": 2.67, "learning_rate": 6.566113222644529e-06, "loss": 0.0824, "step": 8593 }, { "epoch": 2.67, "learning_rate": 6.565713142628526e-06, "loss": 0.0667, "step": 8594 }, { "epoch": 2.67, "learning_rate": 6.565313062612522e-06, "loss": 0.0674, "step": 8595 }, { "epoch": 2.67, "learning_rate": 6.56491298259652e-06, "loss": 0.0423, "step": 8596 }, { "epoch": 2.67, "learning_rate": 6.5645129025805166e-06, "loss": 0.1402, "step": 8597 }, { "epoch": 2.67, "learning_rate": 6.564112822564513e-06, "loss": 0.0457, "step": 8598 }, { "epoch": 2.67, "learning_rate": 6.56371274254851e-06, "loss": 0.0627, "step": 8599 }, { "epoch": 2.67, "learning_rate": 6.5633126625325075e-06, "loss": 0.049, "step": 8600 }, { "epoch": 2.67, "learning_rate": 6.562912582516504e-06, "loss": 0.1101, "step": 8601 }, { "epoch": 2.67, "learning_rate": 6.562512502500501e-06, "loss": 0.0489, "step": 8602 }, { "epoch": 2.67, "learning_rate": 6.5621124224844975e-06, "loss": 0.0856, "step": 8603 }, { "epoch": 2.67, "learning_rate": 6.561712342468495e-06, "loss": 0.137, "step": 8604 }, { "epoch": 2.67, "learning_rate": 6.561312262452491e-06, "loss": 0.0357, "step": 8605 }, { "epoch": 2.67, "learning_rate": 6.560912182436488e-06, "loss": 0.0785, "step": 8606 }, { "epoch": 2.67, "learning_rate": 6.560512102420484e-06, "loss": 0.0495, "step": 8607 }, { "epoch": 2.67, "learning_rate": 6.560112022404481e-06, "loss": 0.0609, "step": 8608 }, { "epoch": 2.67, "learning_rate": 6.5597119423884785e-06, "loss": 0.0971, "step": 8609 }, { "epoch": 2.67, "learning_rate": 6.559311862372475e-06, "loss": 0.0824, "step": 8610 }, { "epoch": 2.67, "learning_rate": 6.558911782356472e-06, "loss": 0.0536, "step": 8611 }, { "epoch": 2.67, "learning_rate": 6.5585117023404686e-06, "loss": 0.1138, "step": 8612 }, { "epoch": 2.67, "learning_rate": 6.558111622324465e-06, "loss": 0.0686, "step": 8613 }, { "epoch": 2.67, "learning_rate": 6.557711542308463e-06, "loss": 0.0486, "step": 8614 }, { "epoch": 2.67, "learning_rate": 6.5573114622924595e-06, "loss": 0.0905, "step": 8615 }, { "epoch": 2.67, "learning_rate": 6.556911382276456e-06, "loss": 0.0893, "step": 8616 }, { "epoch": 2.67, "learning_rate": 6.556511302260453e-06, "loss": 0.2166, "step": 8617 }, { "epoch": 2.67, "learning_rate": 6.556111222244449e-06, "loss": 0.152, "step": 8618 }, { "epoch": 2.68, "learning_rate": 6.555711142228446e-06, "loss": 0.0894, "step": 8619 }, { "epoch": 2.68, "learning_rate": 6.555311062212443e-06, "loss": 0.0939, "step": 8620 }, { "epoch": 2.68, "learning_rate": 6.55491098219644e-06, "loss": 0.1369, "step": 8621 }, { "epoch": 2.68, "learning_rate": 6.554510902180436e-06, "loss": 0.044, "step": 8622 }, { "epoch": 2.68, "learning_rate": 6.554110822164434e-06, "loss": 0.2514, "step": 8623 }, { "epoch": 2.68, "learning_rate": 6.5537107421484305e-06, "loss": 0.1065, "step": 8624 }, { "epoch": 2.68, "learning_rate": 6.553310662132427e-06, "loss": 0.0653, "step": 8625 }, { "epoch": 2.68, "learning_rate": 6.552910582116424e-06, "loss": 0.0493, "step": 8626 }, { "epoch": 2.68, "learning_rate": 6.5525105021004206e-06, "loss": 0.0897, "step": 8627 }, { "epoch": 2.68, "learning_rate": 6.552110422084418e-06, "loss": 0.1311, "step": 8628 }, { "epoch": 2.68, "learning_rate": 6.551710342068415e-06, "loss": 0.0553, "step": 8629 }, { "epoch": 2.68, "learning_rate": 6.551310262052411e-06, "loss": 0.0726, "step": 8630 }, { "epoch": 2.68, "learning_rate": 6.550910182036407e-06, "loss": 0.1147, "step": 8631 }, { "epoch": 2.68, "learning_rate": 6.550510102020404e-06, "loss": 0.0581, "step": 8632 }, { "epoch": 2.68, "learning_rate": 6.5501100220044015e-06, "loss": 0.0857, "step": 8633 }, { "epoch": 2.68, "learning_rate": 6.549709941988398e-06, "loss": 0.0786, "step": 8634 }, { "epoch": 2.68, "learning_rate": 6.549309861972395e-06, "loss": 0.064, "step": 8635 }, { "epoch": 2.68, "learning_rate": 6.5489097819563916e-06, "loss": 0.1048, "step": 8636 }, { "epoch": 2.68, "learning_rate": 6.548509701940389e-06, "loss": 0.0753, "step": 8637 }, { "epoch": 2.68, "learning_rate": 6.548109621924386e-06, "loss": 0.0331, "step": 8638 }, { "epoch": 2.68, "learning_rate": 6.5477095419083825e-06, "loss": 0.0671, "step": 8639 }, { "epoch": 2.68, "learning_rate": 6.547309461892379e-06, "loss": 0.0448, "step": 8640 }, { "epoch": 2.68, "learning_rate": 6.546909381876376e-06, "loss": 0.1168, "step": 8641 }, { "epoch": 2.68, "learning_rate": 6.546509301860373e-06, "loss": 0.1518, "step": 8642 }, { "epoch": 2.68, "learning_rate": 6.546109221844369e-06, "loss": 0.0886, "step": 8643 }, { "epoch": 2.68, "learning_rate": 6.545709141828366e-06, "loss": 0.0964, "step": 8644 }, { "epoch": 2.68, "learning_rate": 6.545309061812363e-06, "loss": 0.0832, "step": 8645 }, { "epoch": 2.68, "learning_rate": 6.544908981796359e-06, "loss": 0.0618, "step": 8646 }, { "epoch": 2.68, "learning_rate": 6.544508901780357e-06, "loss": 0.0758, "step": 8647 }, { "epoch": 2.68, "learning_rate": 6.5441088217643535e-06, "loss": 0.118, "step": 8648 }, { "epoch": 2.68, "learning_rate": 6.54370874174835e-06, "loss": 0.07, "step": 8649 }, { "epoch": 2.68, "learning_rate": 6.543308661732347e-06, "loss": 0.0581, "step": 8650 }, { "epoch": 2.69, "learning_rate": 6.542908581716344e-06, "loss": 0.0864, "step": 8651 }, { "epoch": 2.69, "learning_rate": 6.542508501700341e-06, "loss": 0.1854, "step": 8652 }, { "epoch": 2.69, "learning_rate": 6.542108421684338e-06, "loss": 0.0753, "step": 8653 }, { "epoch": 2.69, "learning_rate": 6.5417083416683345e-06, "loss": 0.1068, "step": 8654 }, { "epoch": 2.69, "learning_rate": 6.54130826165233e-06, "loss": 0.0551, "step": 8655 }, { "epoch": 2.69, "learning_rate": 6.540908181636327e-06, "loss": 0.0815, "step": 8656 }, { "epoch": 2.69, "learning_rate": 6.5405081016203245e-06, "loss": 0.0257, "step": 8657 }, { "epoch": 2.69, "learning_rate": 6.540108021604321e-06, "loss": 0.0572, "step": 8658 }, { "epoch": 2.69, "learning_rate": 6.539707941588318e-06, "loss": 0.0643, "step": 8659 }, { "epoch": 2.69, "learning_rate": 6.539307861572315e-06, "loss": 0.1332, "step": 8660 }, { "epoch": 2.69, "learning_rate": 6.538907781556312e-06, "loss": 0.1284, "step": 8661 }, { "epoch": 2.69, "learning_rate": 6.538507701540309e-06, "loss": 0.0329, "step": 8662 }, { "epoch": 2.69, "learning_rate": 6.5381076215243055e-06, "loss": 0.1034, "step": 8663 }, { "epoch": 2.69, "learning_rate": 6.537707541508302e-06, "loss": 0.0807, "step": 8664 }, { "epoch": 2.69, "learning_rate": 6.5373074614923e-06, "loss": 0.076, "step": 8665 }, { "epoch": 2.69, "learning_rate": 6.536907381476296e-06, "loss": 0.0721, "step": 8666 }, { "epoch": 2.69, "learning_rate": 6.536507301460293e-06, "loss": 0.0908, "step": 8667 }, { "epoch": 2.69, "learning_rate": 6.536107221444289e-06, "loss": 0.1625, "step": 8668 }, { "epoch": 2.69, "learning_rate": 6.535707141428286e-06, "loss": 0.0882, "step": 8669 }, { "epoch": 2.69, "learning_rate": 6.535307061412282e-06, "loss": 0.0565, "step": 8670 }, { "epoch": 2.69, "learning_rate": 6.53490698139628e-06, "loss": 0.0947, "step": 8671 }, { "epoch": 2.69, "learning_rate": 6.5345069013802765e-06, "loss": 0.1252, "step": 8672 }, { "epoch": 2.69, "learning_rate": 6.534106821364273e-06, "loss": 0.0656, "step": 8673 }, { "epoch": 2.69, "learning_rate": 6.53370674134827e-06, "loss": 0.0787, "step": 8674 }, { "epoch": 2.69, "learning_rate": 6.533306661332267e-06, "loss": 0.0431, "step": 8675 }, { "epoch": 2.69, "learning_rate": 6.532906581316264e-06, "loss": 0.0384, "step": 8676 }, { "epoch": 2.69, "learning_rate": 6.532506501300261e-06, "loss": 0.101, "step": 8677 }, { "epoch": 2.69, "learning_rate": 6.5321064212842575e-06, "loss": 0.0472, "step": 8678 }, { "epoch": 2.69, "learning_rate": 6.531706341268255e-06, "loss": 0.0972, "step": 8679 }, { "epoch": 2.69, "learning_rate": 6.531306261252252e-06, "loss": 0.0757, "step": 8680 }, { "epoch": 2.69, "learning_rate": 6.5309061812362475e-06, "loss": 0.082, "step": 8681 }, { "epoch": 2.69, "learning_rate": 6.530506101220244e-06, "loss": 0.1014, "step": 8682 }, { "epoch": 2.7, "learning_rate": 6.530106021204241e-06, "loss": 0.0504, "step": 8683 }, { "epoch": 2.7, "learning_rate": 6.529705941188238e-06, "loss": 0.0942, "step": 8684 }, { "epoch": 2.7, "learning_rate": 6.529305861172235e-06, "loss": 0.1067, "step": 8685 }, { "epoch": 2.7, "learning_rate": 6.528905781156232e-06, "loss": 0.0309, "step": 8686 }, { "epoch": 2.7, "learning_rate": 6.5285057011402285e-06, "loss": 0.0421, "step": 8687 }, { "epoch": 2.7, "learning_rate": 6.528105621124225e-06, "loss": 0.114, "step": 8688 }, { "epoch": 2.7, "learning_rate": 6.527705541108223e-06, "loss": 0.08, "step": 8689 }, { "epoch": 2.7, "learning_rate": 6.527305461092219e-06, "loss": 0.1009, "step": 8690 }, { "epoch": 2.7, "learning_rate": 6.526905381076216e-06, "loss": 0.1033, "step": 8691 }, { "epoch": 2.7, "learning_rate": 6.526505301060213e-06, "loss": 0.0991, "step": 8692 }, { "epoch": 2.7, "learning_rate": 6.526105221044209e-06, "loss": 0.0718, "step": 8693 }, { "epoch": 2.7, "learning_rate": 6.525705141028206e-06, "loss": 0.0406, "step": 8694 }, { "epoch": 2.7, "learning_rate": 6.525305061012203e-06, "loss": 0.0397, "step": 8695 }, { "epoch": 2.7, "learning_rate": 6.5249049809961995e-06, "loss": 0.0568, "step": 8696 }, { "epoch": 2.7, "learning_rate": 6.524504900980196e-06, "loss": 0.1113, "step": 8697 }, { "epoch": 2.7, "learning_rate": 6.524104820964193e-06, "loss": 0.0755, "step": 8698 }, { "epoch": 2.7, "learning_rate": 6.5237047409481904e-06, "loss": 0.0939, "step": 8699 }, { "epoch": 2.7, "learning_rate": 6.523304660932187e-06, "loss": 0.079, "step": 8700 }, { "epoch": 2.7, "learning_rate": 6.522904580916184e-06, "loss": 0.1164, "step": 8701 }, { "epoch": 2.7, "learning_rate": 6.5225045009001805e-06, "loss": 0.0642, "step": 8702 }, { "epoch": 2.7, "learning_rate": 6.522104420884178e-06, "loss": 0.0949, "step": 8703 }, { "epoch": 2.7, "learning_rate": 6.521704340868175e-06, "loss": 0.0644, "step": 8704 }, { "epoch": 2.7, "learning_rate": 6.521304260852171e-06, "loss": 0.1922, "step": 8705 }, { "epoch": 2.7, "learning_rate": 6.520904180836167e-06, "loss": 0.0765, "step": 8706 }, { "epoch": 2.7, "learning_rate": 6.520504100820164e-06, "loss": 0.0687, "step": 8707 }, { "epoch": 2.7, "learning_rate": 6.5201040208041615e-06, "loss": 0.0835, "step": 8708 }, { "epoch": 2.7, "learning_rate": 6.519703940788158e-06, "loss": 0.0727, "step": 8709 }, { "epoch": 2.7, "learning_rate": 6.519303860772155e-06, "loss": 0.1867, "step": 8710 }, { "epoch": 2.7, "learning_rate": 6.5189037807561515e-06, "loss": 0.0504, "step": 8711 }, { "epoch": 2.7, "learning_rate": 6.518503700740148e-06, "loss": 0.046, "step": 8712 }, { "epoch": 2.7, "learning_rate": 6.518103620724146e-06, "loss": 0.058, "step": 8713 }, { "epoch": 2.7, "learning_rate": 6.517703540708142e-06, "loss": 0.0543, "step": 8714 }, { "epoch": 2.7, "learning_rate": 6.517303460692139e-06, "loss": 0.0462, "step": 8715 }, { "epoch": 2.71, "learning_rate": 6.516903380676136e-06, "loss": 0.0792, "step": 8716 }, { "epoch": 2.71, "learning_rate": 6.516503300660133e-06, "loss": 0.2718, "step": 8717 }, { "epoch": 2.71, "learning_rate": 6.516103220644129e-06, "loss": 0.0772, "step": 8718 }, { "epoch": 2.71, "learning_rate": 6.515703140628126e-06, "loss": 0.0648, "step": 8719 }, { "epoch": 2.71, "learning_rate": 6.5153030606121225e-06, "loss": 0.0781, "step": 8720 }, { "epoch": 2.71, "learning_rate": 6.514902980596119e-06, "loss": 0.1332, "step": 8721 }, { "epoch": 2.71, "learning_rate": 6.514502900580117e-06, "loss": 0.0708, "step": 8722 }, { "epoch": 2.71, "learning_rate": 6.5141028205641134e-06, "loss": 0.1377, "step": 8723 }, { "epoch": 2.71, "learning_rate": 6.51370274054811e-06, "loss": 0.0893, "step": 8724 }, { "epoch": 2.71, "learning_rate": 6.513302660532107e-06, "loss": 0.0678, "step": 8725 }, { "epoch": 2.71, "learning_rate": 6.5129025805161035e-06, "loss": 0.1269, "step": 8726 }, { "epoch": 2.71, "learning_rate": 6.512502500500101e-06, "loss": 0.0793, "step": 8727 }, { "epoch": 2.71, "learning_rate": 6.512102420484098e-06, "loss": 0.0598, "step": 8728 }, { "epoch": 2.71, "learning_rate": 6.511702340468094e-06, "loss": 0.1168, "step": 8729 }, { "epoch": 2.71, "learning_rate": 6.511302260452091e-06, "loss": 0.0901, "step": 8730 }, { "epoch": 2.71, "learning_rate": 6.510902180436087e-06, "loss": 0.1124, "step": 8731 }, { "epoch": 2.71, "learning_rate": 6.5105021004200845e-06, "loss": 0.0414, "step": 8732 }, { "epoch": 2.71, "learning_rate": 6.510102020404081e-06, "loss": 0.0608, "step": 8733 }, { "epoch": 2.71, "learning_rate": 6.509701940388078e-06, "loss": 0.0819, "step": 8734 }, { "epoch": 2.71, "learning_rate": 6.5093018603720745e-06, "loss": 0.0896, "step": 8735 }, { "epoch": 2.71, "learning_rate": 6.508901780356072e-06, "loss": 0.0514, "step": 8736 }, { "epoch": 2.71, "learning_rate": 6.508501700340069e-06, "loss": 0.0873, "step": 8737 }, { "epoch": 2.71, "learning_rate": 6.5081016203240654e-06, "loss": 0.0502, "step": 8738 }, { "epoch": 2.71, "learning_rate": 6.507701540308062e-06, "loss": 0.0399, "step": 8739 }, { "epoch": 2.71, "learning_rate": 6.507301460292059e-06, "loss": 0.1776, "step": 8740 }, { "epoch": 2.71, "learning_rate": 6.506901380276056e-06, "loss": 0.0497, "step": 8741 }, { "epoch": 2.71, "learning_rate": 6.506501300260053e-06, "loss": 0.0631, "step": 8742 }, { "epoch": 2.71, "learning_rate": 6.50610122024405e-06, "loss": 0.0352, "step": 8743 }, { "epoch": 2.71, "learning_rate": 6.5057011402280455e-06, "loss": 0.0875, "step": 8744 }, { "epoch": 2.71, "learning_rate": 6.505301060212042e-06, "loss": 0.168, "step": 8745 }, { "epoch": 2.71, "learning_rate": 6.50490098019604e-06, "loss": 0.0978, "step": 8746 }, { "epoch": 2.71, "learning_rate": 6.5045009001800365e-06, "loss": 0.0731, "step": 8747 }, { "epoch": 2.72, "learning_rate": 6.504100820164033e-06, "loss": 0.1768, "step": 8748 }, { "epoch": 2.72, "learning_rate": 6.50370074014803e-06, "loss": 0.0546, "step": 8749 }, { "epoch": 2.72, "learning_rate": 6.503300660132027e-06, "loss": 0.0814, "step": 8750 }, { "epoch": 2.72, "learning_rate": 6.502900580116024e-06, "loss": 0.0266, "step": 8751 }, { "epoch": 2.72, "learning_rate": 6.502500500100021e-06, "loss": 0.1254, "step": 8752 }, { "epoch": 2.72, "learning_rate": 6.502100420084017e-06, "loss": 0.0898, "step": 8753 }, { "epoch": 2.72, "learning_rate": 6.501700340068015e-06, "loss": 0.0698, "step": 8754 }, { "epoch": 2.72, "learning_rate": 6.501300260052012e-06, "loss": 0.0387, "step": 8755 }, { "epoch": 2.72, "learning_rate": 6.5009001800360075e-06, "loss": 0.1834, "step": 8756 }, { "epoch": 2.72, "learning_rate": 6.500500100020004e-06, "loss": 0.0722, "step": 8757 }, { "epoch": 2.72, "learning_rate": 6.500100020004001e-06, "loss": 0.0462, "step": 8758 }, { "epoch": 2.72, "learning_rate": 6.4996999399879975e-06, "loss": 0.1813, "step": 8759 }, { "epoch": 2.72, "learning_rate": 6.499299859971995e-06, "loss": 0.0476, "step": 8760 }, { "epoch": 2.72, "learning_rate": 6.498899779955992e-06, "loss": 0.0961, "step": 8761 }, { "epoch": 2.72, "learning_rate": 6.4984996999399884e-06, "loss": 0.089, "step": 8762 }, { "epoch": 2.72, "learning_rate": 6.498099619923985e-06, "loss": 0.0626, "step": 8763 }, { "epoch": 2.72, "learning_rate": 6.497699539907983e-06, "loss": 0.0875, "step": 8764 }, { "epoch": 2.72, "learning_rate": 6.497299459891979e-06, "loss": 0.0554, "step": 8765 }, { "epoch": 2.72, "learning_rate": 6.496899379875976e-06, "loss": 0.076, "step": 8766 }, { "epoch": 2.72, "learning_rate": 6.496499299859973e-06, "loss": 0.0696, "step": 8767 }, { "epoch": 2.72, "learning_rate": 6.49609921984397e-06, "loss": 0.1128, "step": 8768 }, { "epoch": 2.72, "learning_rate": 6.495699139827965e-06, "loss": 0.1482, "step": 8769 }, { "epoch": 2.72, "learning_rate": 6.495299059811963e-06, "loss": 0.0883, "step": 8770 }, { "epoch": 2.72, "learning_rate": 6.4948989797959595e-06, "loss": 0.1311, "step": 8771 }, { "epoch": 2.72, "learning_rate": 6.494498899779956e-06, "loss": 0.0737, "step": 8772 }, { "epoch": 2.72, "learning_rate": 6.494098819763953e-06, "loss": 0.0886, "step": 8773 }, { "epoch": 2.72, "learning_rate": 6.49369873974795e-06, "loss": 0.1159, "step": 8774 }, { "epoch": 2.72, "learning_rate": 6.493298659731947e-06, "loss": 0.0596, "step": 8775 }, { "epoch": 2.72, "learning_rate": 6.492898579715944e-06, "loss": 0.0843, "step": 8776 }, { "epoch": 2.72, "learning_rate": 6.4924984996999404e-06, "loss": 0.0544, "step": 8777 }, { "epoch": 2.72, "learning_rate": 6.492098419683938e-06, "loss": 0.0946, "step": 8778 }, { "epoch": 2.72, "learning_rate": 6.491698339667935e-06, "loss": 0.0451, "step": 8779 }, { "epoch": 2.73, "learning_rate": 6.491298259651931e-06, "loss": 0.0924, "step": 8780 }, { "epoch": 2.73, "learning_rate": 6.490898179635927e-06, "loss": 0.0624, "step": 8781 }, { "epoch": 2.73, "learning_rate": 6.490498099619924e-06, "loss": 0.0663, "step": 8782 }, { "epoch": 2.73, "learning_rate": 6.4900980196039205e-06, "loss": 0.0767, "step": 8783 }, { "epoch": 2.73, "learning_rate": 6.489697939587918e-06, "loss": 0.1282, "step": 8784 }, { "epoch": 2.73, "learning_rate": 6.489297859571915e-06, "loss": 0.1971, "step": 8785 }, { "epoch": 2.73, "learning_rate": 6.4888977795559115e-06, "loss": 0.0884, "step": 8786 }, { "epoch": 2.73, "learning_rate": 6.488497699539908e-06, "loss": 0.0892, "step": 8787 }, { "epoch": 2.73, "learning_rate": 6.488097619523906e-06, "loss": 0.2222, "step": 8788 }, { "epoch": 2.73, "learning_rate": 6.487697539507902e-06, "loss": 0.086, "step": 8789 }, { "epoch": 2.73, "learning_rate": 6.487297459491899e-06, "loss": 0.0911, "step": 8790 }, { "epoch": 2.73, "learning_rate": 6.486897379475896e-06, "loss": 0.0524, "step": 8791 }, { "epoch": 2.73, "learning_rate": 6.486497299459893e-06, "loss": 0.0822, "step": 8792 }, { "epoch": 2.73, "learning_rate": 6.48609721944389e-06, "loss": 0.1166, "step": 8793 }, { "epoch": 2.73, "learning_rate": 6.485697139427886e-06, "loss": 0.0592, "step": 8794 }, { "epoch": 2.73, "learning_rate": 6.4852970594118825e-06, "loss": 0.0225, "step": 8795 }, { "epoch": 2.73, "learning_rate": 6.484896979395879e-06, "loss": 0.1173, "step": 8796 }, { "epoch": 2.73, "learning_rate": 6.484496899379876e-06, "loss": 0.0718, "step": 8797 }, { "epoch": 2.73, "learning_rate": 6.484096819363873e-06, "loss": 0.0874, "step": 8798 }, { "epoch": 2.73, "learning_rate": 6.48369673934787e-06, "loss": 0.052, "step": 8799 }, { "epoch": 2.73, "learning_rate": 6.483296659331867e-06, "loss": 0.1012, "step": 8800 }, { "epoch": 2.73, "learning_rate": 6.4828965793158634e-06, "loss": 0.0723, "step": 8801 }, { "epoch": 2.73, "learning_rate": 6.482496499299861e-06, "loss": 0.0733, "step": 8802 }, { "epoch": 2.73, "learning_rate": 6.482096419283858e-06, "loss": 0.0702, "step": 8803 }, { "epoch": 2.73, "learning_rate": 6.481696339267854e-06, "loss": 0.1068, "step": 8804 }, { "epoch": 2.73, "learning_rate": 6.481296259251851e-06, "loss": 0.0518, "step": 8805 }, { "epoch": 2.73, "learning_rate": 6.4808961792358486e-06, "loss": 0.0491, "step": 8806 }, { "epoch": 2.73, "learning_rate": 6.480496099219844e-06, "loss": 0.0938, "step": 8807 }, { "epoch": 2.73, "learning_rate": 6.480096019203841e-06, "loss": 0.0765, "step": 8808 }, { "epoch": 2.73, "learning_rate": 6.479695939187838e-06, "loss": 0.0898, "step": 8809 }, { "epoch": 2.73, "learning_rate": 6.4792958591718345e-06, "loss": 0.0682, "step": 8810 }, { "epoch": 2.73, "learning_rate": 6.478895779155831e-06, "loss": 0.0758, "step": 8811 }, { "epoch": 2.74, "learning_rate": 6.478495699139829e-06, "loss": 0.217, "step": 8812 }, { "epoch": 2.74, "learning_rate": 6.478095619123825e-06, "loss": 0.0683, "step": 8813 }, { "epoch": 2.74, "learning_rate": 6.477695539107822e-06, "loss": 0.1602, "step": 8814 }, { "epoch": 2.74, "learning_rate": 6.477295459091819e-06, "loss": 0.1808, "step": 8815 }, { "epoch": 2.74, "learning_rate": 6.476895379075816e-06, "loss": 0.0791, "step": 8816 }, { "epoch": 2.74, "learning_rate": 6.476495299059813e-06, "loss": 0.1371, "step": 8817 }, { "epoch": 2.74, "learning_rate": 6.47609521904381e-06, "loss": 0.061, "step": 8818 }, { "epoch": 2.74, "learning_rate": 6.4756951390278055e-06, "loss": 0.0687, "step": 8819 }, { "epoch": 2.74, "learning_rate": 6.475295059011802e-06, "loss": 0.1637, "step": 8820 }, { "epoch": 2.74, "learning_rate": 6.4748949789958e-06, "loss": 0.0656, "step": 8821 }, { "epoch": 2.74, "learning_rate": 6.474494898979796e-06, "loss": 0.0552, "step": 8822 }, { "epoch": 2.74, "learning_rate": 6.474094818963793e-06, "loss": 0.032, "step": 8823 }, { "epoch": 2.74, "learning_rate": 6.47369473894779e-06, "loss": 0.1535, "step": 8824 }, { "epoch": 2.74, "learning_rate": 6.4732946589317865e-06, "loss": 0.0596, "step": 8825 }, { "epoch": 2.74, "learning_rate": 6.472894578915784e-06, "loss": 0.1396, "step": 8826 }, { "epoch": 2.74, "learning_rate": 6.472494498899781e-06, "loss": 0.0327, "step": 8827 }, { "epoch": 2.74, "learning_rate": 6.472094418883777e-06, "loss": 0.041, "step": 8828 }, { "epoch": 2.74, "learning_rate": 6.471694338867774e-06, "loss": 0.0872, "step": 8829 }, { "epoch": 2.74, "learning_rate": 6.471294258851772e-06, "loss": 0.1072, "step": 8830 }, { "epoch": 2.74, "learning_rate": 6.470894178835768e-06, "loss": 0.0621, "step": 8831 }, { "epoch": 2.74, "learning_rate": 6.470494098819764e-06, "loss": 0.1644, "step": 8832 }, { "epoch": 2.74, "learning_rate": 6.470094018803761e-06, "loss": 0.0511, "step": 8833 }, { "epoch": 2.74, "learning_rate": 6.4696939387877575e-06, "loss": 0.1411, "step": 8834 }, { "epoch": 2.74, "learning_rate": 6.469293858771755e-06, "loss": 0.0456, "step": 8835 }, { "epoch": 2.74, "learning_rate": 6.468893778755752e-06, "loss": 0.0738, "step": 8836 }, { "epoch": 2.74, "learning_rate": 6.468493698739748e-06, "loss": 0.0858, "step": 8837 }, { "epoch": 2.74, "learning_rate": 6.468093618723745e-06, "loss": 0.0782, "step": 8838 }, { "epoch": 2.74, "learning_rate": 6.467693538707743e-06, "loss": 0.0653, "step": 8839 }, { "epoch": 2.74, "learning_rate": 6.467293458691739e-06, "loss": 0.0693, "step": 8840 }, { "epoch": 2.74, "learning_rate": 6.466893378675736e-06, "loss": 0.1214, "step": 8841 }, { "epoch": 2.74, "learning_rate": 6.466493298659733e-06, "loss": 0.0949, "step": 8842 }, { "epoch": 2.74, "learning_rate": 6.466093218643729e-06, "loss": 0.0731, "step": 8843 }, { "epoch": 2.74, "learning_rate": 6.465693138627727e-06, "loss": 0.1008, "step": 8844 }, { "epoch": 2.75, "learning_rate": 6.465293058611723e-06, "loss": 0.1002, "step": 8845 }, { "epoch": 2.75, "learning_rate": 6.464892978595719e-06, "loss": 0.078, "step": 8846 }, { "epoch": 2.75, "learning_rate": 6.464492898579716e-06, "loss": 0.1222, "step": 8847 }, { "epoch": 2.75, "learning_rate": 6.464092818563713e-06, "loss": 0.0339, "step": 8848 }, { "epoch": 2.75, "learning_rate": 6.46369273854771e-06, "loss": 0.0606, "step": 8849 }, { "epoch": 2.75, "learning_rate": 6.463292658531707e-06, "loss": 0.023, "step": 8850 }, { "epoch": 2.75, "learning_rate": 6.462892578515704e-06, "loss": 0.0538, "step": 8851 }, { "epoch": 2.75, "learning_rate": 6.4624924984997e-06, "loss": 0.091, "step": 8852 }, { "epoch": 2.75, "learning_rate": 6.462092418483698e-06, "loss": 0.0862, "step": 8853 }, { "epoch": 2.75, "learning_rate": 6.461692338467695e-06, "loss": 0.0612, "step": 8854 }, { "epoch": 2.75, "learning_rate": 6.461292258451691e-06, "loss": 0.107, "step": 8855 }, { "epoch": 2.75, "learning_rate": 6.460892178435688e-06, "loss": 0.0786, "step": 8856 }, { "epoch": 2.75, "learning_rate": 6.460492098419684e-06, "loss": 0.0907, "step": 8857 }, { "epoch": 2.75, "learning_rate": 6.4600920184036805e-06, "loss": 0.0777, "step": 8858 }, { "epoch": 2.75, "learning_rate": 6.459691938387678e-06, "loss": 0.1048, "step": 8859 }, { "epoch": 2.75, "learning_rate": 6.459291858371675e-06, "loss": 0.0798, "step": 8860 }, { "epoch": 2.75, "learning_rate": 6.458891778355671e-06, "loss": 0.0909, "step": 8861 }, { "epoch": 2.75, "learning_rate": 6.458491698339668e-06, "loss": 0.0834, "step": 8862 }, { "epoch": 2.75, "learning_rate": 6.458091618323666e-06, "loss": 0.0644, "step": 8863 }, { "epoch": 2.75, "learning_rate": 6.457691538307662e-06, "loss": 0.08, "step": 8864 }, { "epoch": 2.75, "learning_rate": 6.457291458291659e-06, "loss": 0.1006, "step": 8865 }, { "epoch": 2.75, "learning_rate": 6.456891378275656e-06, "loss": 0.1234, "step": 8866 }, { "epoch": 2.75, "learning_rate": 6.456491298259653e-06, "loss": 0.0775, "step": 8867 }, { "epoch": 2.75, "learning_rate": 6.45609121824365e-06, "loss": 0.03, "step": 8868 }, { "epoch": 2.75, "learning_rate": 6.455691138227647e-06, "loss": 0.2173, "step": 8869 }, { "epoch": 2.75, "learning_rate": 6.455291058211642e-06, "loss": 0.0633, "step": 8870 }, { "epoch": 2.75, "learning_rate": 6.454890978195639e-06, "loss": 0.1136, "step": 8871 }, { "epoch": 2.75, "learning_rate": 6.454490898179636e-06, "loss": 0.0999, "step": 8872 }, { "epoch": 2.75, "learning_rate": 6.454090818163633e-06, "loss": 0.0865, "step": 8873 }, { "epoch": 2.75, "learning_rate": 6.45369073814763e-06, "loss": 0.0376, "step": 8874 }, { "epoch": 2.75, "learning_rate": 6.453290658131627e-06, "loss": 0.0529, "step": 8875 }, { "epoch": 2.75, "learning_rate": 6.452890578115623e-06, "loss": 0.0557, "step": 8876 }, { "epoch": 2.76, "learning_rate": 6.452490498099621e-06, "loss": 0.0399, "step": 8877 }, { "epoch": 2.76, "learning_rate": 6.452090418083618e-06, "loss": 0.067, "step": 8878 }, { "epoch": 2.76, "learning_rate": 6.451690338067614e-06, "loss": 0.0692, "step": 8879 }, { "epoch": 2.76, "learning_rate": 6.451290258051611e-06, "loss": 0.0516, "step": 8880 }, { "epoch": 2.76, "learning_rate": 6.4508901780356085e-06, "loss": 0.104, "step": 8881 }, { "epoch": 2.76, "learning_rate": 6.4504900980196035e-06, "loss": 0.074, "step": 8882 }, { "epoch": 2.76, "learning_rate": 6.450090018003601e-06, "loss": 0.0911, "step": 8883 }, { "epoch": 2.76, "learning_rate": 6.449689937987598e-06, "loss": 0.0573, "step": 8884 }, { "epoch": 2.76, "learning_rate": 6.449289857971594e-06, "loss": 0.1625, "step": 8885 }, { "epoch": 2.76, "learning_rate": 6.448889777955591e-06, "loss": 0.0503, "step": 8886 }, { "epoch": 2.76, "learning_rate": 6.448489697939589e-06, "loss": 0.133, "step": 8887 }, { "epoch": 2.76, "learning_rate": 6.448089617923585e-06, "loss": 0.0465, "step": 8888 }, { "epoch": 2.76, "learning_rate": 6.447689537907582e-06, "loss": 0.0332, "step": 8889 }, { "epoch": 2.76, "learning_rate": 6.447289457891579e-06, "loss": 0.1024, "step": 8890 }, { "epoch": 2.76, "learning_rate": 6.446889377875576e-06, "loss": 0.0219, "step": 8891 }, { "epoch": 2.76, "learning_rate": 6.446489297859573e-06, "loss": 0.1231, "step": 8892 }, { "epoch": 2.76, "learning_rate": 6.44608921784357e-06, "loss": 0.1595, "step": 8893 }, { "epoch": 2.76, "learning_rate": 6.445689137827566e-06, "loss": 0.0931, "step": 8894 }, { "epoch": 2.76, "learning_rate": 6.445289057811562e-06, "loss": 0.0614, "step": 8895 }, { "epoch": 2.76, "learning_rate": 6.444888977795559e-06, "loss": 0.0594, "step": 8896 }, { "epoch": 2.76, "learning_rate": 6.444488897779556e-06, "loss": 0.0806, "step": 8897 }, { "epoch": 2.76, "learning_rate": 6.444088817763553e-06, "loss": 0.0851, "step": 8898 }, { "epoch": 2.76, "learning_rate": 6.44368873774755e-06, "loss": 0.0908, "step": 8899 }, { "epoch": 2.76, "learning_rate": 6.443288657731546e-06, "loss": 0.1913, "step": 8900 }, { "epoch": 2.76, "learning_rate": 6.442888577715544e-06, "loss": 0.157, "step": 8901 }, { "epoch": 2.76, "learning_rate": 6.442488497699541e-06, "loss": 0.0855, "step": 8902 }, { "epoch": 2.76, "learning_rate": 6.442088417683537e-06, "loss": 0.1603, "step": 8903 }, { "epoch": 2.76, "learning_rate": 6.441688337667534e-06, "loss": 0.1009, "step": 8904 }, { "epoch": 2.76, "learning_rate": 6.4412882576515315e-06, "loss": 0.0422, "step": 8905 }, { "epoch": 2.76, "learning_rate": 6.440888177635528e-06, "loss": 0.0923, "step": 8906 }, { "epoch": 2.76, "learning_rate": 6.440488097619525e-06, "loss": 0.1364, "step": 8907 }, { "epoch": 2.76, "learning_rate": 6.440088017603521e-06, "loss": 0.071, "step": 8908 }, { "epoch": 2.77, "learning_rate": 6.439687937587517e-06, "loss": 0.1064, "step": 8909 }, { "epoch": 2.77, "learning_rate": 6.439287857571515e-06, "loss": 0.065, "step": 8910 }, { "epoch": 2.77, "learning_rate": 6.438887777555512e-06, "loss": 0.0435, "step": 8911 }, { "epoch": 2.77, "learning_rate": 6.438487697539508e-06, "loss": 0.0719, "step": 8912 }, { "epoch": 2.77, "learning_rate": 6.438087617523505e-06, "loss": 0.0641, "step": 8913 }, { "epoch": 2.77, "learning_rate": 6.437687537507502e-06, "loss": 0.0563, "step": 8914 }, { "epoch": 2.77, "learning_rate": 6.437287457491499e-06, "loss": 0.1807, "step": 8915 }, { "epoch": 2.77, "learning_rate": 6.436887377475496e-06, "loss": 0.0945, "step": 8916 }, { "epoch": 2.77, "learning_rate": 6.436487297459493e-06, "loss": 0.0569, "step": 8917 }, { "epoch": 2.77, "learning_rate": 6.436087217443489e-06, "loss": 0.0742, "step": 8918 }, { "epoch": 2.77, "learning_rate": 6.435687137427487e-06, "loss": 0.1166, "step": 8919 }, { "epoch": 2.77, "learning_rate": 6.435287057411483e-06, "loss": 0.1221, "step": 8920 }, { "epoch": 2.77, "learning_rate": 6.434886977395479e-06, "loss": 0.0539, "step": 8921 }, { "epoch": 2.77, "learning_rate": 6.434486897379476e-06, "loss": 0.0565, "step": 8922 }, { "epoch": 2.77, "learning_rate": 6.434086817363473e-06, "loss": 0.1955, "step": 8923 }, { "epoch": 2.77, "learning_rate": 6.43368673734747e-06, "loss": 0.1028, "step": 8924 }, { "epoch": 2.77, "learning_rate": 6.433286657331467e-06, "loss": 0.1839, "step": 8925 }, { "epoch": 2.77, "learning_rate": 6.432886577315464e-06, "loss": 0.0808, "step": 8926 }, { "epoch": 2.77, "learning_rate": 6.43248649729946e-06, "loss": 0.0511, "step": 8927 }, { "epoch": 2.77, "learning_rate": 6.432086417283457e-06, "loss": 0.2308, "step": 8928 }, { "epoch": 2.77, "learning_rate": 6.4316863372674545e-06, "loss": 0.0738, "step": 8929 }, { "epoch": 2.77, "learning_rate": 6.431286257251451e-06, "loss": 0.1006, "step": 8930 }, { "epoch": 2.77, "learning_rate": 6.430886177235448e-06, "loss": 0.0946, "step": 8931 }, { "epoch": 2.77, "learning_rate": 6.430486097219445e-06, "loss": 0.2275, "step": 8932 }, { "epoch": 2.77, "learning_rate": 6.4300860172034404e-06, "loss": 0.0658, "step": 8933 }, { "epoch": 2.77, "learning_rate": 6.429685937187438e-06, "loss": 0.0511, "step": 8934 }, { "epoch": 2.77, "learning_rate": 6.429285857171435e-06, "loss": 0.0735, "step": 8935 }, { "epoch": 2.77, "learning_rate": 6.428885777155431e-06, "loss": 0.0829, "step": 8936 }, { "epoch": 2.77, "learning_rate": 6.428485697139428e-06, "loss": 0.0755, "step": 8937 }, { "epoch": 2.77, "learning_rate": 6.4280856171234256e-06, "loss": 0.0436, "step": 8938 }, { "epoch": 2.77, "learning_rate": 6.427685537107422e-06, "loss": 0.1036, "step": 8939 }, { "epoch": 2.77, "learning_rate": 6.427285457091419e-06, "loss": 0.0856, "step": 8940 }, { "epoch": 2.78, "learning_rate": 6.426885377075416e-06, "loss": 0.0794, "step": 8941 }, { "epoch": 2.78, "learning_rate": 6.426485297059412e-06, "loss": 0.0566, "step": 8942 }, { "epoch": 2.78, "learning_rate": 6.42608521704341e-06, "loss": 0.0916, "step": 8943 }, { "epoch": 2.78, "learning_rate": 6.4256851370274065e-06, "loss": 0.085, "step": 8944 }, { "epoch": 2.78, "learning_rate": 6.425285057011402e-06, "loss": 0.0555, "step": 8945 }, { "epoch": 2.78, "learning_rate": 6.424884976995399e-06, "loss": 0.1195, "step": 8946 }, { "epoch": 2.78, "learning_rate": 6.424484896979396e-06, "loss": 0.0467, "step": 8947 }, { "epoch": 2.78, "learning_rate": 6.424084816963393e-06, "loss": 0.0625, "step": 8948 }, { "epoch": 2.78, "learning_rate": 6.42368473694739e-06, "loss": 0.1109, "step": 8949 }, { "epoch": 2.78, "learning_rate": 6.423284656931387e-06, "loss": 0.0634, "step": 8950 }, { "epoch": 2.78, "learning_rate": 6.422884576915383e-06, "loss": 0.0328, "step": 8951 }, { "epoch": 2.78, "learning_rate": 6.422484496899381e-06, "loss": 0.0495, "step": 8952 }, { "epoch": 2.78, "learning_rate": 6.4220844168833775e-06, "loss": 0.0402, "step": 8953 }, { "epoch": 2.78, "learning_rate": 6.421684336867374e-06, "loss": 0.0666, "step": 8954 }, { "epoch": 2.78, "learning_rate": 6.421284256851371e-06, "loss": 0.053, "step": 8955 }, { "epoch": 2.78, "learning_rate": 6.420884176835368e-06, "loss": 0.0677, "step": 8956 }, { "epoch": 2.78, "learning_rate": 6.420484096819365e-06, "loss": 0.1232, "step": 8957 }, { "epoch": 2.78, "learning_rate": 6.420084016803361e-06, "loss": 0.0408, "step": 8958 }, { "epoch": 2.78, "learning_rate": 6.419683936787358e-06, "loss": 0.0417, "step": 8959 }, { "epoch": 2.78, "learning_rate": 6.419283856771354e-06, "loss": 0.0729, "step": 8960 }, { "epoch": 2.78, "learning_rate": 6.418883776755351e-06, "loss": 0.0583, "step": 8961 }, { "epoch": 2.78, "learning_rate": 6.4184836967393486e-06, "loss": 0.0652, "step": 8962 }, { "epoch": 2.78, "learning_rate": 6.418083616723345e-06, "loss": 0.1883, "step": 8963 }, { "epoch": 2.78, "learning_rate": 6.417683536707342e-06, "loss": 0.0879, "step": 8964 }, { "epoch": 2.78, "learning_rate": 6.417283456691339e-06, "loss": 0.1683, "step": 8965 }, { "epoch": 2.78, "learning_rate": 6.416883376675336e-06, "loss": 0.095, "step": 8966 }, { "epoch": 2.78, "learning_rate": 6.416483296659333e-06, "loss": 0.1607, "step": 8967 }, { "epoch": 2.78, "learning_rate": 6.4160832166433295e-06, "loss": 0.116, "step": 8968 }, { "epoch": 2.78, "learning_rate": 6.415683136627326e-06, "loss": 0.0381, "step": 8969 }, { "epoch": 2.78, "learning_rate": 6.415283056611324e-06, "loss": 0.0557, "step": 8970 }, { "epoch": 2.78, "learning_rate": 6.414882976595319e-06, "loss": 0.0785, "step": 8971 }, { "epoch": 2.78, "learning_rate": 6.414482896579316e-06, "loss": 0.0433, "step": 8972 }, { "epoch": 2.79, "learning_rate": 6.414082816563313e-06, "loss": 0.0721, "step": 8973 }, { "epoch": 2.79, "learning_rate": 6.41368273654731e-06, "loss": 0.0666, "step": 8974 }, { "epoch": 2.79, "learning_rate": 6.413282656531306e-06, "loss": 0.0509, "step": 8975 }, { "epoch": 2.79, "learning_rate": 6.412882576515304e-06, "loss": 0.1477, "step": 8976 }, { "epoch": 2.79, "learning_rate": 6.4124824964993006e-06, "loss": 0.3648, "step": 8977 }, { "epoch": 2.79, "learning_rate": 6.412082416483297e-06, "loss": 0.1011, "step": 8978 }, { "epoch": 2.79, "learning_rate": 6.411682336467294e-06, "loss": 0.0967, "step": 8979 }, { "epoch": 2.79, "learning_rate": 6.4112822564512915e-06, "loss": 0.1138, "step": 8980 }, { "epoch": 2.79, "learning_rate": 6.410882176435288e-06, "loss": 0.0423, "step": 8981 }, { "epoch": 2.79, "learning_rate": 6.410482096419285e-06, "loss": 0.0857, "step": 8982 }, { "epoch": 2.79, "learning_rate": 6.410082016403281e-06, "loss": 0.1728, "step": 8983 }, { "epoch": 2.79, "learning_rate": 6.409681936387277e-06, "loss": 0.0586, "step": 8984 }, { "epoch": 2.79, "learning_rate": 6.409281856371274e-06, "loss": 0.067, "step": 8985 }, { "epoch": 2.79, "learning_rate": 6.408881776355272e-06, "loss": 0.0737, "step": 8986 }, { "epoch": 2.79, "learning_rate": 6.408481696339268e-06, "loss": 0.0944, "step": 8987 }, { "epoch": 2.79, "learning_rate": 6.408081616323265e-06, "loss": 0.1524, "step": 8988 }, { "epoch": 2.79, "learning_rate": 6.407681536307262e-06, "loss": 0.0664, "step": 8989 }, { "epoch": 2.79, "learning_rate": 6.407281456291259e-06, "loss": 0.0692, "step": 8990 }, { "epoch": 2.79, "learning_rate": 6.406881376275256e-06, "loss": 0.1348, "step": 8991 }, { "epoch": 2.79, "learning_rate": 6.4064812962592525e-06, "loss": 0.0914, "step": 8992 }, { "epoch": 2.79, "learning_rate": 6.406081216243249e-06, "loss": 0.0752, "step": 8993 }, { "epoch": 2.79, "learning_rate": 6.405681136227247e-06, "loss": 0.0556, "step": 8994 }, { "epoch": 2.79, "learning_rate": 6.4052810562112435e-06, "loss": 0.078, "step": 8995 }, { "epoch": 2.79, "learning_rate": 6.404880976195239e-06, "loss": 0.2611, "step": 8996 }, { "epoch": 2.79, "learning_rate": 6.404480896179236e-06, "loss": 0.0863, "step": 8997 }, { "epoch": 2.79, "learning_rate": 6.404080816163233e-06, "loss": 0.0702, "step": 8998 }, { "epoch": 2.79, "learning_rate": 6.403680736147229e-06, "loss": 0.0565, "step": 8999 }, { "epoch": 2.79, "learning_rate": 6.403280656131227e-06, "loss": 0.0599, "step": 9000 }, { "epoch": 2.79, "learning_rate": 6.4028805761152236e-06, "loss": 0.0943, "step": 9001 }, { "epoch": 2.79, "learning_rate": 6.40248049609922e-06, "loss": 0.1817, "step": 9002 }, { "epoch": 2.79, "learning_rate": 6.402080416083217e-06, "loss": 0.1189, "step": 9003 }, { "epoch": 2.79, "learning_rate": 6.4016803360672145e-06, "loss": 0.0714, "step": 9004 }, { "epoch": 2.79, "learning_rate": 6.401280256051211e-06, "loss": 0.0767, "step": 9005 }, { "epoch": 2.8, "learning_rate": 6.400880176035208e-06, "loss": 0.0759, "step": 9006 }, { "epoch": 2.8, "learning_rate": 6.4004800960192045e-06, "loss": 0.0451, "step": 9007 }, { "epoch": 2.8, "learning_rate": 6.4000800160032e-06, "loss": 0.1044, "step": 9008 }, { "epoch": 2.8, "learning_rate": 6.399679935987198e-06, "loss": 0.2331, "step": 9009 }, { "epoch": 2.8, "learning_rate": 6.399279855971195e-06, "loss": 0.0725, "step": 9010 }, { "epoch": 2.8, "learning_rate": 6.398879775955191e-06, "loss": 0.0996, "step": 9011 }, { "epoch": 2.8, "learning_rate": 6.398479695939188e-06, "loss": 0.0904, "step": 9012 }, { "epoch": 2.8, "learning_rate": 6.398079615923185e-06, "loss": 0.1441, "step": 9013 }, { "epoch": 2.8, "learning_rate": 6.397679535907182e-06, "loss": 0.0837, "step": 9014 }, { "epoch": 2.8, "learning_rate": 6.397279455891179e-06, "loss": 0.0538, "step": 9015 }, { "epoch": 2.8, "learning_rate": 6.3968793758751756e-06, "loss": 0.0707, "step": 9016 }, { "epoch": 2.8, "learning_rate": 6.396479295859172e-06, "loss": 0.137, "step": 9017 }, { "epoch": 2.8, "learning_rate": 6.39607921584317e-06, "loss": 0.1066, "step": 9018 }, { "epoch": 2.8, "learning_rate": 6.3956791358271665e-06, "loss": 0.0823, "step": 9019 }, { "epoch": 2.8, "learning_rate": 6.395279055811163e-06, "loss": 0.0715, "step": 9020 }, { "epoch": 2.8, "learning_rate": 6.394878975795159e-06, "loss": 0.066, "step": 9021 }, { "epoch": 2.8, "learning_rate": 6.394478895779156e-06, "loss": 0.0701, "step": 9022 }, { "epoch": 2.8, "learning_rate": 6.394078815763153e-06, "loss": 0.0615, "step": 9023 }, { "epoch": 2.8, "learning_rate": 6.39367873574715e-06, "loss": 0.1029, "step": 9024 }, { "epoch": 2.8, "learning_rate": 6.393278655731147e-06, "loss": 0.059, "step": 9025 }, { "epoch": 2.8, "learning_rate": 6.392878575715143e-06, "loss": 0.1071, "step": 9026 }, { "epoch": 2.8, "learning_rate": 6.39247849569914e-06, "loss": 0.1626, "step": 9027 }, { "epoch": 2.8, "learning_rate": 6.3920784156831375e-06, "loss": 0.0727, "step": 9028 }, { "epoch": 2.8, "learning_rate": 6.391678335667134e-06, "loss": 0.0467, "step": 9029 }, { "epoch": 2.8, "learning_rate": 6.391278255651131e-06, "loss": 0.0643, "step": 9030 }, { "epoch": 2.8, "learning_rate": 6.3908781756351275e-06, "loss": 0.0888, "step": 9031 }, { "epoch": 2.8, "learning_rate": 6.390478095619125e-06, "loss": 0.0994, "step": 9032 }, { "epoch": 2.8, "learning_rate": 6.390078015603122e-06, "loss": 0.0478, "step": 9033 }, { "epoch": 2.8, "learning_rate": 6.389677935587118e-06, "loss": 0.1531, "step": 9034 }, { "epoch": 2.8, "learning_rate": 6.389277855571114e-06, "loss": 0.0835, "step": 9035 }, { "epoch": 2.8, "learning_rate": 6.388877775555111e-06, "loss": 0.1766, "step": 9036 }, { "epoch": 2.8, "learning_rate": 6.3884776955391085e-06, "loss": 0.0909, "step": 9037 }, { "epoch": 2.81, "learning_rate": 6.388077615523105e-06, "loss": 0.1155, "step": 9038 }, { "epoch": 2.81, "learning_rate": 6.387677535507102e-06, "loss": 0.1456, "step": 9039 }, { "epoch": 2.81, "learning_rate": 6.3872774554910986e-06, "loss": 0.1061, "step": 9040 }, { "epoch": 2.81, "learning_rate": 6.386877375475095e-06, "loss": 0.0665, "step": 9041 }, { "epoch": 2.81, "learning_rate": 6.386477295459093e-06, "loss": 0.077, "step": 9042 }, { "epoch": 2.81, "learning_rate": 6.3860772154430895e-06, "loss": 0.0784, "step": 9043 }, { "epoch": 2.81, "learning_rate": 6.385677135427086e-06, "loss": 0.1304, "step": 9044 }, { "epoch": 2.81, "learning_rate": 6.385277055411083e-06, "loss": 0.2582, "step": 9045 }, { "epoch": 2.81, "learning_rate": 6.384876975395079e-06, "loss": 0.0467, "step": 9046 }, { "epoch": 2.81, "learning_rate": 6.384476895379076e-06, "loss": 0.076, "step": 9047 }, { "epoch": 2.81, "learning_rate": 6.384076815363073e-06, "loss": 0.1205, "step": 9048 }, { "epoch": 2.81, "learning_rate": 6.38367673534707e-06, "loss": 0.213, "step": 9049 }, { "epoch": 2.81, "learning_rate": 6.383276655331066e-06, "loss": 0.0563, "step": 9050 }, { "epoch": 2.81, "learning_rate": 6.382876575315064e-06, "loss": 0.0577, "step": 9051 }, { "epoch": 2.81, "learning_rate": 6.3824764952990605e-06, "loss": 0.0974, "step": 9052 }, { "epoch": 2.81, "learning_rate": 6.382076415283057e-06, "loss": 0.1113, "step": 9053 }, { "epoch": 2.81, "learning_rate": 6.381676335267054e-06, "loss": 0.0447, "step": 9054 }, { "epoch": 2.81, "learning_rate": 6.381276255251051e-06, "loss": 0.1455, "step": 9055 }, { "epoch": 2.81, "learning_rate": 6.380876175235048e-06, "loss": 0.051, "step": 9056 }, { "epoch": 2.81, "learning_rate": 6.380476095219045e-06, "loss": 0.1521, "step": 9057 }, { "epoch": 2.81, "learning_rate": 6.3800760152030415e-06, "loss": 0.0611, "step": 9058 }, { "epoch": 2.81, "learning_rate": 6.379675935187037e-06, "loss": 0.089, "step": 9059 }, { "epoch": 2.81, "learning_rate": 6.379275855171034e-06, "loss": 0.0889, "step": 9060 }, { "epoch": 2.81, "learning_rate": 6.3788757751550315e-06, "loss": 0.0527, "step": 9061 }, { "epoch": 2.81, "learning_rate": 6.378475695139028e-06, "loss": 0.1464, "step": 9062 }, { "epoch": 2.81, "learning_rate": 6.378075615123025e-06, "loss": 0.0683, "step": 9063 }, { "epoch": 2.81, "learning_rate": 6.377675535107022e-06, "loss": 0.0378, "step": 9064 }, { "epoch": 2.81, "learning_rate": 6.377275455091019e-06, "loss": 0.0625, "step": 9065 }, { "epoch": 2.81, "learning_rate": 6.376875375075016e-06, "loss": 0.0348, "step": 9066 }, { "epoch": 2.81, "learning_rate": 6.3764752950590125e-06, "loss": 0.0547, "step": 9067 }, { "epoch": 2.81, "learning_rate": 6.376075215043009e-06, "loss": 0.1025, "step": 9068 }, { "epoch": 2.81, "learning_rate": 6.375675135027007e-06, "loss": 0.042, "step": 9069 }, { "epoch": 2.82, "learning_rate": 6.375275055011003e-06, "loss": 0.0643, "step": 9070 }, { "epoch": 2.82, "learning_rate": 6.374874974995e-06, "loss": 0.0835, "step": 9071 }, { "epoch": 2.82, "learning_rate": 6.374474894978996e-06, "loss": 0.0957, "step": 9072 }, { "epoch": 2.82, "learning_rate": 6.374074814962993e-06, "loss": 0.0599, "step": 9073 }, { "epoch": 2.82, "learning_rate": 6.373674734946989e-06, "loss": 0.051, "step": 9074 }, { "epoch": 2.82, "learning_rate": 6.373274654930987e-06, "loss": 0.0338, "step": 9075 }, { "epoch": 2.82, "learning_rate": 6.3728745749149835e-06, "loss": 0.026, "step": 9076 }, { "epoch": 2.82, "learning_rate": 6.37247449489898e-06, "loss": 0.0961, "step": 9077 }, { "epoch": 2.82, "learning_rate": 6.372074414882977e-06, "loss": 0.1002, "step": 9078 }, { "epoch": 2.82, "learning_rate": 6.371674334866974e-06, "loss": 0.0594, "step": 9079 }, { "epoch": 2.82, "learning_rate": 6.371274254850971e-06, "loss": 0.0776, "step": 9080 }, { "epoch": 2.82, "learning_rate": 6.370874174834968e-06, "loss": 0.059, "step": 9081 }, { "epoch": 2.82, "learning_rate": 6.3704740948189645e-06, "loss": 0.0982, "step": 9082 }, { "epoch": 2.82, "learning_rate": 6.370074014802962e-06, "loss": 0.0759, "step": 9083 }, { "epoch": 2.82, "learning_rate": 6.369673934786957e-06, "loss": 0.1674, "step": 9084 }, { "epoch": 2.82, "learning_rate": 6.3692738547709545e-06, "loss": 0.2882, "step": 9085 }, { "epoch": 2.82, "learning_rate": 6.368873774754951e-06, "loss": 0.0443, "step": 9086 }, { "epoch": 2.82, "learning_rate": 6.368473694738948e-06, "loss": 0.0746, "step": 9087 }, { "epoch": 2.82, "learning_rate": 6.368073614722945e-06, "loss": 0.0876, "step": 9088 }, { "epoch": 2.82, "learning_rate": 6.367673534706942e-06, "loss": 0.0663, "step": 9089 }, { "epoch": 2.82, "learning_rate": 6.367273454690939e-06, "loss": 0.0394, "step": 9090 }, { "epoch": 2.82, "learning_rate": 6.3668733746749355e-06, "loss": 0.2096, "step": 9091 }, { "epoch": 2.82, "learning_rate": 6.366473294658932e-06, "loss": 0.1647, "step": 9092 }, { "epoch": 2.82, "learning_rate": 6.36607321464293e-06, "loss": 0.0733, "step": 9093 }, { "epoch": 2.82, "learning_rate": 6.365673134626926e-06, "loss": 0.1329, "step": 9094 }, { "epoch": 2.82, "learning_rate": 6.365273054610923e-06, "loss": 0.093, "step": 9095 }, { "epoch": 2.82, "learning_rate": 6.36487297459492e-06, "loss": 0.0852, "step": 9096 }, { "epoch": 2.82, "learning_rate": 6.364472894578916e-06, "loss": 0.0962, "step": 9097 }, { "epoch": 2.82, "learning_rate": 6.364072814562912e-06, "loss": 0.1102, "step": 9098 }, { "epoch": 2.82, "learning_rate": 6.36367273454691e-06, "loss": 0.1347, "step": 9099 }, { "epoch": 2.82, "learning_rate": 6.3632726545309065e-06, "loss": 0.1139, "step": 9100 }, { "epoch": 2.82, "learning_rate": 6.362872574514903e-06, "loss": 0.0985, "step": 9101 }, { "epoch": 2.83, "learning_rate": 6.3624724944989e-06, "loss": 0.1423, "step": 9102 }, { "epoch": 2.83, "learning_rate": 6.3620724144828974e-06, "loss": 0.0946, "step": 9103 }, { "epoch": 2.83, "learning_rate": 6.361672334466894e-06, "loss": 0.0873, "step": 9104 }, { "epoch": 2.83, "learning_rate": 6.361272254450891e-06, "loss": 0.0846, "step": 9105 }, { "epoch": 2.83, "learning_rate": 6.3608721744348875e-06, "loss": 0.2348, "step": 9106 }, { "epoch": 2.83, "learning_rate": 6.360472094418885e-06, "loss": 0.1105, "step": 9107 }, { "epoch": 2.83, "learning_rate": 6.360072014402882e-06, "loss": 0.0749, "step": 9108 }, { "epoch": 2.83, "learning_rate": 6.3596719343868775e-06, "loss": 0.0947, "step": 9109 }, { "epoch": 2.83, "learning_rate": 6.359271854370874e-06, "loss": 0.0411, "step": 9110 }, { "epoch": 2.83, "learning_rate": 6.358871774354871e-06, "loss": 0.1007, "step": 9111 }, { "epoch": 2.83, "learning_rate": 6.358471694338868e-06, "loss": 0.0783, "step": 9112 }, { "epoch": 2.83, "learning_rate": 6.358071614322865e-06, "loss": 0.0633, "step": 9113 }, { "epoch": 2.83, "learning_rate": 6.357671534306862e-06, "loss": 0.1299, "step": 9114 }, { "epoch": 2.83, "learning_rate": 6.3572714542908585e-06, "loss": 0.0807, "step": 9115 }, { "epoch": 2.83, "learning_rate": 6.356871374274855e-06, "loss": 0.0657, "step": 9116 }, { "epoch": 2.83, "learning_rate": 6.356471294258853e-06, "loss": 0.086, "step": 9117 }, { "epoch": 2.83, "learning_rate": 6.356071214242849e-06, "loss": 0.09, "step": 9118 }, { "epoch": 2.83, "learning_rate": 6.355671134226846e-06, "loss": 0.0837, "step": 9119 }, { "epoch": 2.83, "learning_rate": 6.355271054210843e-06, "loss": 0.0332, "step": 9120 }, { "epoch": 2.83, "learning_rate": 6.35487097419484e-06, "loss": 0.0755, "step": 9121 }, { "epoch": 2.83, "learning_rate": 6.354470894178836e-06, "loss": 0.1244, "step": 9122 }, { "epoch": 2.83, "learning_rate": 6.354070814162833e-06, "loss": 0.1456, "step": 9123 }, { "epoch": 2.83, "learning_rate": 6.3536707341468295e-06, "loss": 0.0967, "step": 9124 }, { "epoch": 2.83, "learning_rate": 6.353270654130826e-06, "loss": 0.0383, "step": 9125 }, { "epoch": 2.83, "learning_rate": 6.352870574114824e-06, "loss": 0.0834, "step": 9126 }, { "epoch": 2.83, "learning_rate": 6.3524704940988204e-06, "loss": 0.0625, "step": 9127 }, { "epoch": 2.83, "learning_rate": 6.352070414082817e-06, "loss": 0.0902, "step": 9128 }, { "epoch": 2.83, "learning_rate": 6.351670334066814e-06, "loss": 0.1826, "step": 9129 }, { "epoch": 2.83, "learning_rate": 6.3512702540508105e-06, "loss": 0.0525, "step": 9130 }, { "epoch": 2.83, "learning_rate": 6.350870174034808e-06, "loss": 0.1043, "step": 9131 }, { "epoch": 2.83, "learning_rate": 6.350470094018805e-06, "loss": 0.0888, "step": 9132 }, { "epoch": 2.83, "learning_rate": 6.350070014002801e-06, "loss": 0.0675, "step": 9133 }, { "epoch": 2.83, "learning_rate": 6.349669933986798e-06, "loss": 0.1376, "step": 9134 }, { "epoch": 2.84, "learning_rate": 6.349269853970794e-06, "loss": 0.1998, "step": 9135 }, { "epoch": 2.84, "learning_rate": 6.3488697739547915e-06, "loss": 0.112, "step": 9136 }, { "epoch": 2.84, "learning_rate": 6.348469693938788e-06, "loss": 0.1008, "step": 9137 }, { "epoch": 2.84, "learning_rate": 6.348069613922785e-06, "loss": 0.0522, "step": 9138 }, { "epoch": 2.84, "learning_rate": 6.3476695339067815e-06, "loss": 0.0342, "step": 9139 }, { "epoch": 2.84, "learning_rate": 6.347269453890779e-06, "loss": 0.2122, "step": 9140 }, { "epoch": 2.84, "learning_rate": 6.346869373874776e-06, "loss": 0.1535, "step": 9141 }, { "epoch": 2.84, "learning_rate": 6.3464692938587724e-06, "loss": 0.0852, "step": 9142 }, { "epoch": 2.84, "learning_rate": 6.346069213842769e-06, "loss": 0.107, "step": 9143 }, { "epoch": 2.84, "learning_rate": 6.345669133826766e-06, "loss": 0.1357, "step": 9144 }, { "epoch": 2.84, "learning_rate": 6.345269053810763e-06, "loss": 0.0584, "step": 9145 }, { "epoch": 2.84, "learning_rate": 6.34486897379476e-06, "loss": 0.0462, "step": 9146 }, { "epoch": 2.84, "learning_rate": 6.344468893778756e-06, "loss": 0.1115, "step": 9147 }, { "epoch": 2.84, "learning_rate": 6.3440688137627525e-06, "loss": 0.0862, "step": 9148 }, { "epoch": 2.84, "learning_rate": 6.343668733746749e-06, "loss": 0.036, "step": 9149 }, { "epoch": 2.84, "learning_rate": 6.343268653730747e-06, "loss": 0.0697, "step": 9150 }, { "epoch": 2.84, "learning_rate": 6.3428685737147435e-06, "loss": 0.0735, "step": 9151 }, { "epoch": 2.84, "learning_rate": 6.34246849369874e-06, "loss": 0.0411, "step": 9152 }, { "epoch": 2.84, "learning_rate": 6.342068413682737e-06, "loss": 0.0592, "step": 9153 }, { "epoch": 2.84, "learning_rate": 6.341668333666734e-06, "loss": 0.0787, "step": 9154 }, { "epoch": 2.84, "learning_rate": 6.341268253650731e-06, "loss": 0.1684, "step": 9155 }, { "epoch": 2.84, "learning_rate": 6.340868173634728e-06, "loss": 0.0612, "step": 9156 }, { "epoch": 2.84, "learning_rate": 6.340468093618724e-06, "loss": 0.077, "step": 9157 }, { "epoch": 2.84, "learning_rate": 6.340068013602721e-06, "loss": 0.0674, "step": 9158 }, { "epoch": 2.84, "learning_rate": 6.339667933586719e-06, "loss": 0.0748, "step": 9159 }, { "epoch": 2.84, "learning_rate": 6.3392678535707145e-06, "loss": 0.1129, "step": 9160 }, { "epoch": 2.84, "learning_rate": 6.338867773554711e-06, "loss": 0.0945, "step": 9161 }, { "epoch": 2.84, "learning_rate": 6.338467693538708e-06, "loss": 0.1042, "step": 9162 }, { "epoch": 2.84, "learning_rate": 6.3380676135227045e-06, "loss": 0.0861, "step": 9163 }, { "epoch": 2.84, "learning_rate": 6.337667533506702e-06, "loss": 0.057, "step": 9164 }, { "epoch": 2.84, "learning_rate": 6.337267453490699e-06, "loss": 0.0641, "step": 9165 }, { "epoch": 2.84, "learning_rate": 6.3368673734746954e-06, "loss": 0.1305, "step": 9166 }, { "epoch": 2.85, "learning_rate": 6.336467293458692e-06, "loss": 0.0963, "step": 9167 }, { "epoch": 2.85, "learning_rate": 6.33606721344269e-06, "loss": 0.0557, "step": 9168 }, { "epoch": 2.85, "learning_rate": 6.335667133426686e-06, "loss": 0.0237, "step": 9169 }, { "epoch": 2.85, "learning_rate": 6.335267053410683e-06, "loss": 0.1468, "step": 9170 }, { "epoch": 2.85, "learning_rate": 6.33486697339468e-06, "loss": 0.1512, "step": 9171 }, { "epoch": 2.85, "learning_rate": 6.3344668933786756e-06, "loss": 0.1555, "step": 9172 }, { "epoch": 2.85, "learning_rate": 6.334066813362672e-06, "loss": 0.095, "step": 9173 }, { "epoch": 2.85, "learning_rate": 6.33366673334667e-06, "loss": 0.1277, "step": 9174 }, { "epoch": 2.85, "learning_rate": 6.3332666533306665e-06, "loss": 0.2583, "step": 9175 }, { "epoch": 2.85, "learning_rate": 6.332866573314663e-06, "loss": 0.0606, "step": 9176 }, { "epoch": 2.85, "learning_rate": 6.33246649329866e-06, "loss": 0.0381, "step": 9177 }, { "epoch": 2.85, "learning_rate": 6.332066413282657e-06, "loss": 0.1282, "step": 9178 }, { "epoch": 2.85, "learning_rate": 6.331666333266654e-06, "loss": 0.1225, "step": 9179 }, { "epoch": 2.85, "learning_rate": 6.331266253250651e-06, "loss": 0.0767, "step": 9180 }, { "epoch": 2.85, "learning_rate": 6.3308661732346474e-06, "loss": 0.1103, "step": 9181 }, { "epoch": 2.85, "learning_rate": 6.330466093218645e-06, "loss": 0.0663, "step": 9182 }, { "epoch": 2.85, "learning_rate": 6.330066013202642e-06, "loss": 0.0366, "step": 9183 }, { "epoch": 2.85, "learning_rate": 6.329665933186638e-06, "loss": 0.0983, "step": 9184 }, { "epoch": 2.85, "learning_rate": 6.329265853170634e-06, "loss": 0.0922, "step": 9185 }, { "epoch": 2.85, "learning_rate": 6.328865773154631e-06, "loss": 0.1861, "step": 9186 }, { "epoch": 2.85, "learning_rate": 6.3284656931386275e-06, "loss": 0.1007, "step": 9187 }, { "epoch": 2.85, "learning_rate": 6.328065613122625e-06, "loss": 0.1067, "step": 9188 }, { "epoch": 2.85, "learning_rate": 6.327665533106622e-06, "loss": 0.1459, "step": 9189 }, { "epoch": 2.85, "learning_rate": 6.3272654530906185e-06, "loss": 0.0512, "step": 9190 }, { "epoch": 2.85, "learning_rate": 6.326865373074615e-06, "loss": 0.0495, "step": 9191 }, { "epoch": 2.85, "learning_rate": 6.326465293058613e-06, "loss": 0.0829, "step": 9192 }, { "epoch": 2.85, "learning_rate": 6.326065213042609e-06, "loss": 0.0515, "step": 9193 }, { "epoch": 2.85, "learning_rate": 6.325665133026606e-06, "loss": 0.1305, "step": 9194 }, { "epoch": 2.85, "learning_rate": 6.325265053010603e-06, "loss": 0.0504, "step": 9195 }, { "epoch": 2.85, "learning_rate": 6.3248649729946e-06, "loss": 0.1063, "step": 9196 }, { "epoch": 2.85, "learning_rate": 6.324464892978597e-06, "loss": 0.0685, "step": 9197 }, { "epoch": 2.85, "learning_rate": 6.324064812962593e-06, "loss": 0.1105, "step": 9198 }, { "epoch": 2.86, "learning_rate": 6.3236647329465895e-06, "loss": 0.0631, "step": 9199 }, { "epoch": 2.86, "learning_rate": 6.323264652930586e-06, "loss": 0.0932, "step": 9200 }, { "epoch": 2.86, "learning_rate": 6.322864572914583e-06, "loss": 0.3287, "step": 9201 }, { "epoch": 2.86, "learning_rate": 6.32246449289858e-06, "loss": 0.1023, "step": 9202 }, { "epoch": 2.86, "learning_rate": 6.322064412882577e-06, "loss": 0.1187, "step": 9203 }, { "epoch": 2.86, "learning_rate": 6.321664332866574e-06, "loss": 0.1521, "step": 9204 }, { "epoch": 2.86, "learning_rate": 6.3212642528505704e-06, "loss": 0.0809, "step": 9205 }, { "epoch": 2.86, "learning_rate": 6.320864172834568e-06, "loss": 0.0475, "step": 9206 }, { "epoch": 2.86, "learning_rate": 6.320464092818565e-06, "loss": 0.0666, "step": 9207 }, { "epoch": 2.86, "learning_rate": 6.320064012802561e-06, "loss": 0.1269, "step": 9208 }, { "epoch": 2.86, "learning_rate": 6.319663932786558e-06, "loss": 0.1218, "step": 9209 }, { "epoch": 2.86, "learning_rate": 6.319263852770554e-06, "loss": 0.0481, "step": 9210 }, { "epoch": 2.86, "learning_rate": 6.318863772754551e-06, "loss": 0.1787, "step": 9211 }, { "epoch": 2.86, "learning_rate": 6.318463692738548e-06, "loss": 0.0911, "step": 9212 }, { "epoch": 2.86, "learning_rate": 6.318063612722545e-06, "loss": 0.0433, "step": 9213 }, { "epoch": 2.86, "learning_rate": 6.3176635327065415e-06, "loss": 0.048, "step": 9214 }, { "epoch": 2.86, "learning_rate": 6.317263452690538e-06, "loss": 0.1509, "step": 9215 }, { "epoch": 2.86, "learning_rate": 6.316863372674536e-06, "loss": 0.062, "step": 9216 }, { "epoch": 2.86, "learning_rate": 6.316463292658532e-06, "loss": 0.0773, "step": 9217 }, { "epoch": 2.86, "learning_rate": 6.316063212642529e-06, "loss": 0.0675, "step": 9218 }, { "epoch": 2.86, "learning_rate": 6.315663132626526e-06, "loss": 0.0614, "step": 9219 }, { "epoch": 2.86, "learning_rate": 6.315263052610523e-06, "loss": 0.0418, "step": 9220 }, { "epoch": 2.86, "learning_rate": 6.31486297259452e-06, "loss": 0.0871, "step": 9221 }, { "epoch": 2.86, "learning_rate": 6.314462892578517e-06, "loss": 0.0412, "step": 9222 }, { "epoch": 2.86, "learning_rate": 6.3140628125625125e-06, "loss": 0.1104, "step": 9223 }, { "epoch": 2.86, "learning_rate": 6.313662732546509e-06, "loss": 0.0734, "step": 9224 }, { "epoch": 2.86, "learning_rate": 6.313262652530507e-06, "loss": 0.1488, "step": 9225 }, { "epoch": 2.86, "learning_rate": 6.312862572514503e-06, "loss": 0.0487, "step": 9226 }, { "epoch": 2.86, "learning_rate": 6.3124624924985e-06, "loss": 0.0637, "step": 9227 }, { "epoch": 2.86, "learning_rate": 6.312062412482497e-06, "loss": 0.1271, "step": 9228 }, { "epoch": 2.86, "learning_rate": 6.3116623324664935e-06, "loss": 0.0537, "step": 9229 }, { "epoch": 2.86, "learning_rate": 6.311262252450491e-06, "loss": 0.0995, "step": 9230 }, { "epoch": 2.87, "learning_rate": 6.310862172434488e-06, "loss": 0.0893, "step": 9231 }, { "epoch": 2.87, "learning_rate": 6.310462092418484e-06, "loss": 0.0525, "step": 9232 }, { "epoch": 2.87, "learning_rate": 6.310062012402481e-06, "loss": 0.095, "step": 9233 }, { "epoch": 2.87, "learning_rate": 6.309661932386479e-06, "loss": 0.0908, "step": 9234 }, { "epoch": 2.87, "learning_rate": 6.309261852370474e-06, "loss": 0.0536, "step": 9235 }, { "epoch": 2.87, "learning_rate": 6.308861772354471e-06, "loss": 0.0513, "step": 9236 }, { "epoch": 2.87, "learning_rate": 6.308461692338468e-06, "loss": 0.0697, "step": 9237 }, { "epoch": 2.87, "learning_rate": 6.3080616123224645e-06, "loss": 0.0634, "step": 9238 }, { "epoch": 2.87, "learning_rate": 6.307661532306462e-06, "loss": 0.0651, "step": 9239 }, { "epoch": 2.87, "learning_rate": 6.307261452290459e-06, "loss": 0.0332, "step": 9240 }, { "epoch": 2.87, "learning_rate": 6.306861372274455e-06, "loss": 0.2286, "step": 9241 }, { "epoch": 2.87, "learning_rate": 6.306461292258452e-06, "loss": 0.0725, "step": 9242 }, { "epoch": 2.87, "learning_rate": 6.306061212242449e-06, "loss": 0.1983, "step": 9243 }, { "epoch": 2.87, "learning_rate": 6.305661132226446e-06, "loss": 0.0819, "step": 9244 }, { "epoch": 2.87, "learning_rate": 6.305261052210443e-06, "loss": 0.0904, "step": 9245 }, { "epoch": 2.87, "learning_rate": 6.30486097219444e-06, "loss": 0.1175, "step": 9246 }, { "epoch": 2.87, "learning_rate": 6.304460892178436e-06, "loss": 0.0674, "step": 9247 }, { "epoch": 2.87, "learning_rate": 6.304060812162432e-06, "loss": 0.0814, "step": 9248 }, { "epoch": 2.87, "learning_rate": 6.30366073214643e-06, "loss": 0.0181, "step": 9249 }, { "epoch": 2.87, "learning_rate": 6.303260652130426e-06, "loss": 0.0802, "step": 9250 }, { "epoch": 2.87, "learning_rate": 6.302860572114423e-06, "loss": 0.1332, "step": 9251 }, { "epoch": 2.87, "learning_rate": 6.30246049209842e-06, "loss": 0.0703, "step": 9252 }, { "epoch": 2.87, "learning_rate": 6.302060412082417e-06, "loss": 0.057, "step": 9253 }, { "epoch": 2.87, "learning_rate": 6.301660332066414e-06, "loss": 0.11, "step": 9254 }, { "epoch": 2.87, "learning_rate": 6.301260252050411e-06, "loss": 0.0846, "step": 9255 }, { "epoch": 2.87, "learning_rate": 6.300860172034407e-06, "loss": 0.0713, "step": 9256 }, { "epoch": 2.87, "learning_rate": 6.300460092018404e-06, "loss": 0.098, "step": 9257 }, { "epoch": 2.87, "learning_rate": 6.300060012002402e-06, "loss": 0.0715, "step": 9258 }, { "epoch": 2.87, "learning_rate": 6.299659931986398e-06, "loss": 0.1295, "step": 9259 }, { "epoch": 2.87, "learning_rate": 6.299259851970395e-06, "loss": 0.0823, "step": 9260 }, { "epoch": 2.87, "learning_rate": 6.298859771954391e-06, "loss": 0.078, "step": 9261 }, { "epoch": 2.87, "learning_rate": 6.2984596919383875e-06, "loss": 0.089, "step": 9262 }, { "epoch": 2.88, "learning_rate": 6.298059611922385e-06, "loss": 0.0476, "step": 9263 }, { "epoch": 2.88, "learning_rate": 6.297659531906382e-06, "loss": 0.0737, "step": 9264 }, { "epoch": 2.88, "learning_rate": 6.297259451890378e-06, "loss": 0.2588, "step": 9265 }, { "epoch": 2.88, "learning_rate": 6.296859371874375e-06, "loss": 0.0422, "step": 9266 }, { "epoch": 2.88, "learning_rate": 6.296459291858373e-06, "loss": 0.0964, "step": 9267 }, { "epoch": 2.88, "learning_rate": 6.296059211842369e-06, "loss": 0.1059, "step": 9268 }, { "epoch": 2.88, "learning_rate": 6.295659131826366e-06, "loss": 0.1382, "step": 9269 }, { "epoch": 2.88, "learning_rate": 6.295259051810363e-06, "loss": 0.0442, "step": 9270 }, { "epoch": 2.88, "learning_rate": 6.29485897179436e-06, "loss": 0.0606, "step": 9271 }, { "epoch": 2.88, "learning_rate": 6.294458891778357e-06, "loss": 0.0573, "step": 9272 }, { "epoch": 2.88, "learning_rate": 6.294058811762353e-06, "loss": 0.0709, "step": 9273 }, { "epoch": 2.88, "learning_rate": 6.293658731746349e-06, "loss": 0.0343, "step": 9274 }, { "epoch": 2.88, "learning_rate": 6.293258651730346e-06, "loss": 0.0673, "step": 9275 }, { "epoch": 2.88, "learning_rate": 6.292858571714343e-06, "loss": 0.1175, "step": 9276 }, { "epoch": 2.88, "learning_rate": 6.29245849169834e-06, "loss": 0.071, "step": 9277 }, { "epoch": 2.88, "learning_rate": 6.292058411682337e-06, "loss": 0.0599, "step": 9278 }, { "epoch": 2.88, "learning_rate": 6.291658331666334e-06, "loss": 0.0656, "step": 9279 }, { "epoch": 2.88, "learning_rate": 6.29125825165033e-06, "loss": 0.0415, "step": 9280 }, { "epoch": 2.88, "learning_rate": 6.290858171634328e-06, "loss": 0.0654, "step": 9281 }, { "epoch": 2.88, "learning_rate": 6.290458091618325e-06, "loss": 0.0945, "step": 9282 }, { "epoch": 2.88, "learning_rate": 6.290058011602321e-06, "loss": 0.0868, "step": 9283 }, { "epoch": 2.88, "learning_rate": 6.289657931586318e-06, "loss": 0.0956, "step": 9284 }, { "epoch": 2.88, "learning_rate": 6.2892578515703155e-06, "loss": 0.0435, "step": 9285 }, { "epoch": 2.88, "learning_rate": 6.2888577715543105e-06, "loss": 0.0823, "step": 9286 }, { "epoch": 2.88, "learning_rate": 6.288457691538308e-06, "loss": 0.0777, "step": 9287 }, { "epoch": 2.88, "learning_rate": 6.288057611522305e-06, "loss": 0.0809, "step": 9288 }, { "epoch": 2.88, "learning_rate": 6.287657531506301e-06, "loss": 0.0415, "step": 9289 }, { "epoch": 2.88, "learning_rate": 6.287257451490298e-06, "loss": 0.1235, "step": 9290 }, { "epoch": 2.88, "learning_rate": 6.286857371474296e-06, "loss": 0.0539, "step": 9291 }, { "epoch": 2.88, "learning_rate": 6.286457291458292e-06, "loss": 0.1609, "step": 9292 }, { "epoch": 2.88, "learning_rate": 6.286057211442289e-06, "loss": 0.1028, "step": 9293 }, { "epoch": 2.88, "learning_rate": 6.285657131426286e-06, "loss": 0.1821, "step": 9294 }, { "epoch": 2.88, "learning_rate": 6.285257051410283e-06, "loss": 0.1174, "step": 9295 }, { "epoch": 2.89, "learning_rate": 6.28485697139428e-06, "loss": 0.0827, "step": 9296 }, { "epoch": 2.89, "learning_rate": 6.284456891378277e-06, "loss": 0.0814, "step": 9297 }, { "epoch": 2.89, "learning_rate": 6.284056811362273e-06, "loss": 0.0287, "step": 9298 }, { "epoch": 2.89, "learning_rate": 6.283656731346269e-06, "loss": 0.1018, "step": 9299 }, { "epoch": 2.89, "learning_rate": 6.283256651330266e-06, "loss": 0.2099, "step": 9300 }, { "epoch": 2.89, "learning_rate": 6.282856571314263e-06, "loss": 0.1315, "step": 9301 }, { "epoch": 2.89, "learning_rate": 6.28245649129826e-06, "loss": 0.096, "step": 9302 }, { "epoch": 2.89, "learning_rate": 6.282056411282257e-06, "loss": 0.078, "step": 9303 }, { "epoch": 2.89, "learning_rate": 6.281656331266253e-06, "loss": 0.0395, "step": 9304 }, { "epoch": 2.89, "learning_rate": 6.281256251250251e-06, "loss": 0.0761, "step": 9305 }, { "epoch": 2.89, "learning_rate": 6.280856171234248e-06, "loss": 0.121, "step": 9306 }, { "epoch": 2.89, "learning_rate": 6.280456091218244e-06, "loss": 0.1291, "step": 9307 }, { "epoch": 2.89, "learning_rate": 6.280056011202241e-06, "loss": 0.0506, "step": 9308 }, { "epoch": 2.89, "learning_rate": 6.2796559311862385e-06, "loss": 0.2146, "step": 9309 }, { "epoch": 2.89, "learning_rate": 6.279255851170235e-06, "loss": 0.0899, "step": 9310 }, { "epoch": 2.89, "learning_rate": 6.278855771154231e-06, "loss": 0.0949, "step": 9311 }, { "epoch": 2.89, "learning_rate": 6.278455691138228e-06, "loss": 0.1199, "step": 9312 }, { "epoch": 2.89, "learning_rate": 6.278055611122224e-06, "loss": 0.0632, "step": 9313 }, { "epoch": 2.89, "learning_rate": 6.277655531106221e-06, "loss": 0.1107, "step": 9314 }, { "epoch": 2.89, "learning_rate": 6.277255451090219e-06, "loss": 0.0674, "step": 9315 }, { "epoch": 2.89, "learning_rate": 6.276855371074215e-06, "loss": 0.0478, "step": 9316 }, { "epoch": 2.89, "learning_rate": 6.276455291058212e-06, "loss": 0.0624, "step": 9317 }, { "epoch": 2.89, "learning_rate": 6.276055211042209e-06, "loss": 0.0933, "step": 9318 }, { "epoch": 2.89, "learning_rate": 6.275655131026206e-06, "loss": 0.0979, "step": 9319 }, { "epoch": 2.89, "learning_rate": 6.275255051010203e-06, "loss": 0.0231, "step": 9320 }, { "epoch": 2.89, "learning_rate": 6.2748549709942e-06, "loss": 0.0515, "step": 9321 }, { "epoch": 2.89, "learning_rate": 6.274454890978196e-06, "loss": 0.0333, "step": 9322 }, { "epoch": 2.89, "learning_rate": 6.274054810962194e-06, "loss": 0.0228, "step": 9323 }, { "epoch": 2.89, "learning_rate": 6.27365473094619e-06, "loss": 0.0611, "step": 9324 }, { "epoch": 2.89, "learning_rate": 6.273254650930186e-06, "loss": 0.0856, "step": 9325 }, { "epoch": 2.89, "learning_rate": 6.272854570914183e-06, "loss": 0.1171, "step": 9326 }, { "epoch": 2.89, "learning_rate": 6.27245449089818e-06, "loss": 0.0852, "step": 9327 }, { "epoch": 2.9, "learning_rate": 6.272054410882176e-06, "loss": 0.0772, "step": 9328 }, { "epoch": 2.9, "learning_rate": 6.271654330866174e-06, "loss": 0.0601, "step": 9329 }, { "epoch": 2.9, "learning_rate": 6.271254250850171e-06, "loss": 0.1049, "step": 9330 }, { "epoch": 2.9, "learning_rate": 6.270854170834167e-06, "loss": 0.1035, "step": 9331 }, { "epoch": 2.9, "learning_rate": 6.270454090818164e-06, "loss": 0.0725, "step": 9332 }, { "epoch": 2.9, "learning_rate": 6.2700540108021615e-06, "loss": 0.0958, "step": 9333 }, { "epoch": 2.9, "learning_rate": 6.269653930786158e-06, "loss": 0.0393, "step": 9334 }, { "epoch": 2.9, "learning_rate": 6.269253850770155e-06, "loss": 0.0563, "step": 9335 }, { "epoch": 2.9, "learning_rate": 6.268853770754151e-06, "loss": 0.047, "step": 9336 }, { "epoch": 2.9, "learning_rate": 6.2684536907381474e-06, "loss": 0.0551, "step": 9337 }, { "epoch": 2.9, "learning_rate": 6.268053610722145e-06, "loss": 0.0655, "step": 9338 }, { "epoch": 2.9, "learning_rate": 6.267653530706142e-06, "loss": 0.0949, "step": 9339 }, { "epoch": 2.9, "learning_rate": 6.267253450690138e-06, "loss": 0.0326, "step": 9340 }, { "epoch": 2.9, "learning_rate": 6.266853370674135e-06, "loss": 0.2315, "step": 9341 }, { "epoch": 2.9, "learning_rate": 6.266453290658132e-06, "loss": 0.0549, "step": 9342 }, { "epoch": 2.9, "learning_rate": 6.266053210642129e-06, "loss": 0.0723, "step": 9343 }, { "epoch": 2.9, "learning_rate": 6.265653130626126e-06, "loss": 0.0374, "step": 9344 }, { "epoch": 2.9, "learning_rate": 6.265253050610123e-06, "loss": 0.0871, "step": 9345 }, { "epoch": 2.9, "learning_rate": 6.264852970594119e-06, "loss": 0.0622, "step": 9346 }, { "epoch": 2.9, "learning_rate": 6.264452890578117e-06, "loss": 0.1063, "step": 9347 }, { "epoch": 2.9, "learning_rate": 6.2640528105621135e-06, "loss": 0.0775, "step": 9348 }, { "epoch": 2.9, "learning_rate": 6.263652730546109e-06, "loss": 0.0891, "step": 9349 }, { "epoch": 2.9, "learning_rate": 6.263252650530106e-06, "loss": 0.1374, "step": 9350 }, { "epoch": 2.9, "learning_rate": 6.262852570514103e-06, "loss": 0.0631, "step": 9351 }, { "epoch": 2.9, "learning_rate": 6.2624524904981e-06, "loss": 0.0866, "step": 9352 }, { "epoch": 2.9, "learning_rate": 6.262052410482097e-06, "loss": 0.1934, "step": 9353 }, { "epoch": 2.9, "learning_rate": 6.261652330466094e-06, "loss": 0.0939, "step": 9354 }, { "epoch": 2.9, "learning_rate": 6.26125225045009e-06, "loss": 0.1287, "step": 9355 }, { "epoch": 2.9, "learning_rate": 6.260852170434088e-06, "loss": 0.0989, "step": 9356 }, { "epoch": 2.9, "learning_rate": 6.2604520904180845e-06, "loss": 0.1347, "step": 9357 }, { "epoch": 2.9, "learning_rate": 6.260052010402081e-06, "loss": 0.1001, "step": 9358 }, { "epoch": 2.9, "learning_rate": 6.259651930386078e-06, "loss": 0.184, "step": 9359 }, { "epoch": 2.91, "learning_rate": 6.259251850370075e-06, "loss": 0.0795, "step": 9360 }, { "epoch": 2.91, "learning_rate": 6.258851770354072e-06, "loss": 0.027, "step": 9361 }, { "epoch": 2.91, "learning_rate": 6.258451690338068e-06, "loss": 0.0497, "step": 9362 }, { "epoch": 2.91, "learning_rate": 6.258051610322065e-06, "loss": 0.0697, "step": 9363 }, { "epoch": 2.91, "learning_rate": 6.257651530306061e-06, "loss": 0.0921, "step": 9364 }, { "epoch": 2.91, "learning_rate": 6.257251450290058e-06, "loss": 0.1687, "step": 9365 }, { "epoch": 2.91, "learning_rate": 6.2568513702740556e-06, "loss": 0.0458, "step": 9366 }, { "epoch": 2.91, "learning_rate": 6.256451290258052e-06, "loss": 0.0856, "step": 9367 }, { "epoch": 2.91, "learning_rate": 6.256051210242049e-06, "loss": 0.0959, "step": 9368 }, { "epoch": 2.91, "learning_rate": 6.255651130226046e-06, "loss": 0.0293, "step": 9369 }, { "epoch": 2.91, "learning_rate": 6.255251050210043e-06, "loss": 0.1097, "step": 9370 }, { "epoch": 2.91, "learning_rate": 6.25485097019404e-06, "loss": 0.0351, "step": 9371 }, { "epoch": 2.91, "learning_rate": 6.2544508901780365e-06, "loss": 0.0784, "step": 9372 }, { "epoch": 2.91, "learning_rate": 6.254050810162033e-06, "loss": 0.1246, "step": 9373 }, { "epoch": 2.91, "learning_rate": 6.253650730146029e-06, "loss": 0.0752, "step": 9374 }, { "epoch": 2.91, "learning_rate": 6.253250650130026e-06, "loss": 0.0536, "step": 9375 }, { "epoch": 2.91, "learning_rate": 6.252850570114023e-06, "loss": 0.0838, "step": 9376 }, { "epoch": 2.91, "learning_rate": 6.25245049009802e-06, "loss": 0.0786, "step": 9377 }, { "epoch": 2.91, "learning_rate": 6.252050410082017e-06, "loss": 0.0717, "step": 9378 }, { "epoch": 2.91, "learning_rate": 6.251650330066013e-06, "loss": 0.0628, "step": 9379 }, { "epoch": 2.91, "learning_rate": 6.251250250050011e-06, "loss": 0.1348, "step": 9380 }, { "epoch": 2.91, "learning_rate": 6.2508501700340076e-06, "loss": 0.0545, "step": 9381 }, { "epoch": 2.91, "learning_rate": 6.250450090018004e-06, "loss": 0.0594, "step": 9382 }, { "epoch": 2.91, "learning_rate": 6.250050010002001e-06, "loss": 0.0654, "step": 9383 }, { "epoch": 2.91, "learning_rate": 6.2496499299859985e-06, "loss": 0.07, "step": 9384 }, { "epoch": 2.91, "learning_rate": 6.249249849969995e-06, "loss": 0.124, "step": 9385 }, { "epoch": 2.91, "learning_rate": 6.248849769953992e-06, "loss": 0.1082, "step": 9386 }, { "epoch": 2.91, "learning_rate": 6.248449689937988e-06, "loss": 0.0489, "step": 9387 }, { "epoch": 2.91, "learning_rate": 6.248049609921984e-06, "loss": 0.1212, "step": 9388 }, { "epoch": 2.91, "learning_rate": 6.247649529905981e-06, "loss": 0.0478, "step": 9389 }, { "epoch": 2.91, "learning_rate": 6.2472494498899786e-06, "loss": 0.1128, "step": 9390 }, { "epoch": 2.91, "learning_rate": 6.246849369873975e-06, "loss": 0.0585, "step": 9391 }, { "epoch": 2.92, "learning_rate": 6.246449289857972e-06, "loss": 0.0664, "step": 9392 }, { "epoch": 2.92, "learning_rate": 6.246049209841969e-06, "loss": 0.1432, "step": 9393 }, { "epoch": 2.92, "learning_rate": 6.245649129825966e-06, "loss": 0.0551, "step": 9394 }, { "epoch": 2.92, "learning_rate": 6.245249049809963e-06, "loss": 0.0643, "step": 9395 }, { "epoch": 2.92, "learning_rate": 6.2448489697939595e-06, "loss": 0.0682, "step": 9396 }, { "epoch": 2.92, "learning_rate": 6.244448889777956e-06, "loss": 0.1911, "step": 9397 }, { "epoch": 2.92, "learning_rate": 6.244048809761954e-06, "loss": 0.0654, "step": 9398 }, { "epoch": 2.92, "learning_rate": 6.243648729745949e-06, "loss": 0.0743, "step": 9399 }, { "epoch": 2.92, "learning_rate": 6.243248649729946e-06, "loss": 0.08, "step": 9400 }, { "epoch": 2.92, "learning_rate": 6.242848569713943e-06, "loss": 0.0588, "step": 9401 }, { "epoch": 2.92, "learning_rate": 6.24244848969794e-06, "loss": 0.0737, "step": 9402 }, { "epoch": 2.92, "learning_rate": 6.242048409681936e-06, "loss": 0.0639, "step": 9403 }, { "epoch": 2.92, "learning_rate": 6.241648329665934e-06, "loss": 0.0459, "step": 9404 }, { "epoch": 2.92, "learning_rate": 6.2412482496499306e-06, "loss": 0.1205, "step": 9405 }, { "epoch": 2.92, "learning_rate": 6.240848169633927e-06, "loss": 0.1701, "step": 9406 }, { "epoch": 2.92, "learning_rate": 6.240448089617924e-06, "loss": 0.2021, "step": 9407 }, { "epoch": 2.92, "learning_rate": 6.2400480096019215e-06, "loss": 0.0688, "step": 9408 }, { "epoch": 2.92, "learning_rate": 6.239647929585918e-06, "loss": 0.1392, "step": 9409 }, { "epoch": 2.92, "learning_rate": 6.239247849569915e-06, "loss": 0.074, "step": 9410 }, { "epoch": 2.92, "learning_rate": 6.2388477695539115e-06, "loss": 0.197, "step": 9411 }, { "epoch": 2.92, "learning_rate": 6.238447689537907e-06, "loss": 0.0857, "step": 9412 }, { "epoch": 2.92, "learning_rate": 6.238047609521904e-06, "loss": 0.0359, "step": 9413 }, { "epoch": 2.92, "learning_rate": 6.237647529505902e-06, "loss": 0.0849, "step": 9414 }, { "epoch": 2.92, "learning_rate": 6.237247449489898e-06, "loss": 0.1175, "step": 9415 }, { "epoch": 2.92, "learning_rate": 6.236847369473895e-06, "loss": 0.0842, "step": 9416 }, { "epoch": 2.92, "learning_rate": 6.236447289457892e-06, "loss": 0.063, "step": 9417 }, { "epoch": 2.92, "learning_rate": 6.236047209441889e-06, "loss": 0.1997, "step": 9418 }, { "epoch": 2.92, "learning_rate": 6.235647129425886e-06, "loss": 0.0971, "step": 9419 }, { "epoch": 2.92, "learning_rate": 6.2352470494098826e-06, "loss": 0.0589, "step": 9420 }, { "epoch": 2.92, "learning_rate": 6.234846969393879e-06, "loss": 0.1127, "step": 9421 }, { "epoch": 2.92, "learning_rate": 6.234446889377877e-06, "loss": 0.1044, "step": 9422 }, { "epoch": 2.92, "learning_rate": 6.2340468093618735e-06, "loss": 0.0357, "step": 9423 }, { "epoch": 2.93, "learning_rate": 6.23364672934587e-06, "loss": 0.0245, "step": 9424 }, { "epoch": 2.93, "learning_rate": 6.233246649329866e-06, "loss": 0.2237, "step": 9425 }, { "epoch": 2.93, "learning_rate": 6.232846569313863e-06, "loss": 0.0285, "step": 9426 }, { "epoch": 2.93, "learning_rate": 6.23244648929786e-06, "loss": 0.0266, "step": 9427 }, { "epoch": 2.93, "learning_rate": 6.232046409281857e-06, "loss": 0.0473, "step": 9428 }, { "epoch": 2.93, "learning_rate": 6.2316463292658536e-06, "loss": 0.1305, "step": 9429 }, { "epoch": 2.93, "learning_rate": 6.23124624924985e-06, "loss": 0.0912, "step": 9430 }, { "epoch": 2.93, "learning_rate": 6.230846169233847e-06, "loss": 0.0645, "step": 9431 }, { "epoch": 2.93, "learning_rate": 6.2304460892178445e-06, "loss": 0.1394, "step": 9432 }, { "epoch": 2.93, "learning_rate": 6.230046009201841e-06, "loss": 0.2402, "step": 9433 }, { "epoch": 2.93, "learning_rate": 6.229645929185838e-06, "loss": 0.0753, "step": 9434 }, { "epoch": 2.93, "learning_rate": 6.2292458491698345e-06, "loss": 0.0408, "step": 9435 }, { "epoch": 2.93, "learning_rate": 6.228845769153832e-06, "loss": 0.063, "step": 9436 }, { "epoch": 2.93, "learning_rate": 6.228445689137828e-06, "loss": 0.0528, "step": 9437 }, { "epoch": 2.93, "learning_rate": 6.228045609121825e-06, "loss": 0.0572, "step": 9438 }, { "epoch": 2.93, "learning_rate": 6.227645529105821e-06, "loss": 0.0658, "step": 9439 }, { "epoch": 2.93, "learning_rate": 6.227245449089818e-06, "loss": 0.0445, "step": 9440 }, { "epoch": 2.93, "learning_rate": 6.2268453690738155e-06, "loss": 0.071, "step": 9441 }, { "epoch": 2.93, "learning_rate": 6.226445289057812e-06, "loss": 0.1194, "step": 9442 }, { "epoch": 2.93, "learning_rate": 6.226045209041809e-06, "loss": 0.0707, "step": 9443 }, { "epoch": 2.93, "learning_rate": 6.2256451290258056e-06, "loss": 0.0443, "step": 9444 }, { "epoch": 2.93, "learning_rate": 6.225245049009802e-06, "loss": 0.241, "step": 9445 }, { "epoch": 2.93, "learning_rate": 6.2248449689938e-06, "loss": 0.1054, "step": 9446 }, { "epoch": 2.93, "learning_rate": 6.2244448889777965e-06, "loss": 0.085, "step": 9447 }, { "epoch": 2.93, "learning_rate": 6.224044808961793e-06, "loss": 0.0878, "step": 9448 }, { "epoch": 2.93, "learning_rate": 6.22364472894579e-06, "loss": 0.145, "step": 9449 }, { "epoch": 2.93, "learning_rate": 6.223244648929786e-06, "loss": 0.0661, "step": 9450 }, { "epoch": 2.93, "learning_rate": 6.222844568913783e-06, "loss": 0.0372, "step": 9451 }, { "epoch": 2.93, "learning_rate": 6.22244448889778e-06, "loss": 0.1066, "step": 9452 }, { "epoch": 2.93, "learning_rate": 6.222044408881777e-06, "loss": 0.0983, "step": 9453 }, { "epoch": 2.93, "learning_rate": 6.221644328865773e-06, "loss": 0.1143, "step": 9454 }, { "epoch": 2.93, "learning_rate": 6.221244248849771e-06, "loss": 0.0563, "step": 9455 }, { "epoch": 2.93, "learning_rate": 6.2208441688337675e-06, "loss": 0.0323, "step": 9456 }, { "epoch": 2.94, "learning_rate": 6.220444088817764e-06, "loss": 0.1274, "step": 9457 }, { "epoch": 2.94, "learning_rate": 6.220044008801761e-06, "loss": 0.0742, "step": 9458 }, { "epoch": 2.94, "learning_rate": 6.2196439287857576e-06, "loss": 0.0541, "step": 9459 }, { "epoch": 2.94, "learning_rate": 6.219243848769755e-06, "loss": 0.0602, "step": 9460 }, { "epoch": 2.94, "learning_rate": 6.218843768753752e-06, "loss": 0.1925, "step": 9461 }, { "epoch": 2.94, "learning_rate": 6.218443688737748e-06, "loss": 0.0951, "step": 9462 }, { "epoch": 2.94, "learning_rate": 6.218043608721744e-06, "loss": 0.0481, "step": 9463 }, { "epoch": 2.94, "learning_rate": 6.217643528705741e-06, "loss": 0.0412, "step": 9464 }, { "epoch": 2.94, "learning_rate": 6.2172434486897385e-06, "loss": 0.0684, "step": 9465 }, { "epoch": 2.94, "learning_rate": 6.216843368673735e-06, "loss": 0.0652, "step": 9466 }, { "epoch": 2.94, "learning_rate": 6.216443288657732e-06, "loss": 0.0639, "step": 9467 }, { "epoch": 2.94, "learning_rate": 6.2160432086417286e-06, "loss": 0.0842, "step": 9468 }, { "epoch": 2.94, "learning_rate": 6.215643128625726e-06, "loss": 0.1093, "step": 9469 }, { "epoch": 2.94, "learning_rate": 6.215243048609723e-06, "loss": 0.0541, "step": 9470 }, { "epoch": 2.94, "learning_rate": 6.2148429685937195e-06, "loss": 0.1017, "step": 9471 }, { "epoch": 2.94, "learning_rate": 6.214442888577716e-06, "loss": 0.0823, "step": 9472 }, { "epoch": 2.94, "learning_rate": 6.214042808561713e-06, "loss": 0.0773, "step": 9473 }, { "epoch": 2.94, "learning_rate": 6.21364272854571e-06, "loss": 0.0738, "step": 9474 }, { "epoch": 2.94, "learning_rate": 6.213242648529706e-06, "loss": 0.0904, "step": 9475 }, { "epoch": 2.94, "learning_rate": 6.212842568513703e-06, "loss": 0.0531, "step": 9476 }, { "epoch": 2.94, "learning_rate": 6.2124424884977e-06, "loss": 0.1595, "step": 9477 }, { "epoch": 2.94, "learning_rate": 6.212042408481696e-06, "loss": 0.1094, "step": 9478 }, { "epoch": 2.94, "learning_rate": 6.211642328465694e-06, "loss": 0.2074, "step": 9479 }, { "epoch": 2.94, "learning_rate": 6.2112422484496905e-06, "loss": 0.059, "step": 9480 }, { "epoch": 2.94, "learning_rate": 6.210842168433687e-06, "loss": 0.0919, "step": 9481 }, { "epoch": 2.94, "learning_rate": 6.210442088417684e-06, "loss": 0.1065, "step": 9482 }, { "epoch": 2.94, "learning_rate": 6.210042008401681e-06, "loss": 0.0875, "step": 9483 }, { "epoch": 2.94, "learning_rate": 6.209641928385678e-06, "loss": 0.0686, "step": 9484 }, { "epoch": 2.94, "learning_rate": 6.209241848369675e-06, "loss": 0.1043, "step": 9485 }, { "epoch": 2.94, "learning_rate": 6.2088417683536715e-06, "loss": 0.0771, "step": 9486 }, { "epoch": 2.94, "learning_rate": 6.208441688337669e-06, "loss": 0.0894, "step": 9487 }, { "epoch": 2.94, "learning_rate": 6.208041608321664e-06, "loss": 0.0394, "step": 9488 }, { "epoch": 2.95, "learning_rate": 6.2076415283056615e-06, "loss": 0.2443, "step": 9489 }, { "epoch": 2.95, "learning_rate": 6.207241448289658e-06, "loss": 0.0416, "step": 9490 }, { "epoch": 2.95, "learning_rate": 6.206841368273655e-06, "loss": 0.0998, "step": 9491 }, { "epoch": 2.95, "learning_rate": 6.206441288257652e-06, "loss": 0.0765, "step": 9492 }, { "epoch": 2.95, "learning_rate": 6.206041208241649e-06, "loss": 0.0429, "step": 9493 }, { "epoch": 2.95, "learning_rate": 6.205641128225646e-06, "loss": 0.0732, "step": 9494 }, { "epoch": 2.95, "learning_rate": 6.2052410482096425e-06, "loss": 0.091, "step": 9495 }, { "epoch": 2.95, "learning_rate": 6.204840968193639e-06, "loss": 0.0486, "step": 9496 }, { "epoch": 2.95, "learning_rate": 6.204440888177637e-06, "loss": 0.1457, "step": 9497 }, { "epoch": 2.95, "learning_rate": 6.204040808161633e-06, "loss": 0.0373, "step": 9498 }, { "epoch": 2.95, "learning_rate": 6.20364072814563e-06, "loss": 0.1172, "step": 9499 }, { "epoch": 2.95, "learning_rate": 6.203240648129626e-06, "loss": 0.0607, "step": 9500 }, { "epoch": 2.95, "learning_rate": 6.202840568113623e-06, "loss": 0.0568, "step": 9501 }, { "epoch": 2.95, "learning_rate": 6.202440488097619e-06, "loss": 0.1859, "step": 9502 }, { "epoch": 2.95, "learning_rate": 6.202040408081617e-06, "loss": 0.1136, "step": 9503 }, { "epoch": 2.95, "learning_rate": 6.2016403280656135e-06, "loss": 0.1535, "step": 9504 }, { "epoch": 2.95, "learning_rate": 6.20124024804961e-06, "loss": 0.0383, "step": 9505 }, { "epoch": 2.95, "learning_rate": 6.200840168033607e-06, "loss": 0.0593, "step": 9506 }, { "epoch": 2.95, "learning_rate": 6.200440088017604e-06, "loss": 0.0662, "step": 9507 }, { "epoch": 2.95, "learning_rate": 6.200040008001601e-06, "loss": 0.0512, "step": 9508 }, { "epoch": 2.95, "learning_rate": 6.199639927985598e-06, "loss": 0.0487, "step": 9509 }, { "epoch": 2.95, "learning_rate": 6.1992398479695945e-06, "loss": 0.0483, "step": 9510 }, { "epoch": 2.95, "learning_rate": 6.198839767953592e-06, "loss": 0.0641, "step": 9511 }, { "epoch": 2.95, "learning_rate": 6.198439687937589e-06, "loss": 0.1262, "step": 9512 }, { "epoch": 2.95, "learning_rate": 6.1980396079215845e-06, "loss": 0.1217, "step": 9513 }, { "epoch": 2.95, "learning_rate": 6.197639527905581e-06, "loss": 0.0475, "step": 9514 }, { "epoch": 2.95, "learning_rate": 6.197239447889578e-06, "loss": 0.0701, "step": 9515 }, { "epoch": 2.95, "learning_rate": 6.196839367873575e-06, "loss": 0.0794, "step": 9516 }, { "epoch": 2.95, "learning_rate": 6.196439287857572e-06, "loss": 0.0624, "step": 9517 }, { "epoch": 2.95, "learning_rate": 6.196039207841569e-06, "loss": 0.0553, "step": 9518 }, { "epoch": 2.95, "learning_rate": 6.1956391278255655e-06, "loss": 0.0836, "step": 9519 }, { "epoch": 2.95, "learning_rate": 6.195239047809562e-06, "loss": 0.1202, "step": 9520 }, { "epoch": 2.96, "learning_rate": 6.19483896779356e-06, "loss": 0.0508, "step": 9521 }, { "epoch": 2.96, "learning_rate": 6.194438887777556e-06, "loss": 0.1515, "step": 9522 }, { "epoch": 2.96, "learning_rate": 6.194038807761553e-06, "loss": 0.0955, "step": 9523 }, { "epoch": 2.96, "learning_rate": 6.19363872774555e-06, "loss": 0.089, "step": 9524 }, { "epoch": 2.96, "learning_rate": 6.193238647729546e-06, "loss": 0.0504, "step": 9525 }, { "epoch": 2.96, "learning_rate": 6.192838567713543e-06, "loss": 0.0489, "step": 9526 }, { "epoch": 2.96, "learning_rate": 6.19243848769754e-06, "loss": 0.1433, "step": 9527 }, { "epoch": 2.96, "learning_rate": 6.1920384076815365e-06, "loss": 0.0326, "step": 9528 }, { "epoch": 2.96, "learning_rate": 6.191638327665533e-06, "loss": 0.0619, "step": 9529 }, { "epoch": 2.96, "learning_rate": 6.19123824764953e-06, "loss": 0.0695, "step": 9530 }, { "epoch": 2.96, "learning_rate": 6.1908381676335274e-06, "loss": 0.0474, "step": 9531 }, { "epoch": 2.96, "learning_rate": 6.190438087617524e-06, "loss": 0.0745, "step": 9532 }, { "epoch": 2.96, "learning_rate": 6.190038007601521e-06, "loss": 0.1831, "step": 9533 }, { "epoch": 2.96, "learning_rate": 6.1896379275855175e-06, "loss": 0.0838, "step": 9534 }, { "epoch": 2.96, "learning_rate": 6.189237847569515e-06, "loss": 0.0864, "step": 9535 }, { "epoch": 2.96, "learning_rate": 6.188837767553512e-06, "loss": 0.1363, "step": 9536 }, { "epoch": 2.96, "learning_rate": 6.188437687537508e-06, "loss": 0.148, "step": 9537 }, { "epoch": 2.96, "learning_rate": 6.188037607521504e-06, "loss": 0.1595, "step": 9538 }, { "epoch": 2.96, "learning_rate": 6.187637527505501e-06, "loss": 0.0567, "step": 9539 }, { "epoch": 2.96, "learning_rate": 6.1872374474894985e-06, "loss": 0.1718, "step": 9540 }, { "epoch": 2.96, "learning_rate": 6.186837367473495e-06, "loss": 0.1172, "step": 9541 }, { "epoch": 2.96, "learning_rate": 6.186437287457492e-06, "loss": 0.0495, "step": 9542 }, { "epoch": 2.96, "learning_rate": 6.1860372074414885e-06, "loss": 0.1101, "step": 9543 }, { "epoch": 2.96, "learning_rate": 6.185637127425485e-06, "loss": 0.1191, "step": 9544 }, { "epoch": 2.96, "learning_rate": 6.185237047409483e-06, "loss": 0.0667, "step": 9545 }, { "epoch": 2.96, "learning_rate": 6.184836967393479e-06, "loss": 0.0148, "step": 9546 }, { "epoch": 2.96, "learning_rate": 6.184436887377476e-06, "loss": 0.0809, "step": 9547 }, { "epoch": 2.96, "learning_rate": 6.184036807361473e-06, "loss": 0.0763, "step": 9548 }, { "epoch": 2.96, "learning_rate": 6.18363672734547e-06, "loss": 0.0797, "step": 9549 }, { "epoch": 2.96, "learning_rate": 6.183236647329467e-06, "loss": 0.0782, "step": 9550 }, { "epoch": 2.96, "learning_rate": 6.182836567313463e-06, "loss": 0.0826, "step": 9551 }, { "epoch": 2.96, "learning_rate": 6.1824364872974595e-06, "loss": 0.0962, "step": 9552 }, { "epoch": 2.97, "learning_rate": 6.182036407281456e-06, "loss": 0.102, "step": 9553 }, { "epoch": 2.97, "learning_rate": 6.181636327265454e-06, "loss": 0.0478, "step": 9554 }, { "epoch": 2.97, "learning_rate": 6.1812362472494504e-06, "loss": 0.1019, "step": 9555 }, { "epoch": 2.97, "learning_rate": 6.180836167233447e-06, "loss": 0.0846, "step": 9556 }, { "epoch": 2.97, "learning_rate": 6.180436087217444e-06, "loss": 0.0479, "step": 9557 }, { "epoch": 2.97, "learning_rate": 6.1800360072014405e-06, "loss": 0.1178, "step": 9558 }, { "epoch": 2.97, "learning_rate": 6.179635927185438e-06, "loss": 0.085, "step": 9559 }, { "epoch": 2.97, "learning_rate": 6.179235847169435e-06, "loss": 0.1226, "step": 9560 }, { "epoch": 2.97, "learning_rate": 6.178835767153431e-06, "loss": 0.0507, "step": 9561 }, { "epoch": 2.97, "learning_rate": 6.178435687137428e-06, "loss": 0.0965, "step": 9562 }, { "epoch": 2.97, "learning_rate": 6.178035607121424e-06, "loss": 0.0729, "step": 9563 }, { "epoch": 2.97, "learning_rate": 6.1776355271054215e-06, "loss": 0.0781, "step": 9564 }, { "epoch": 2.97, "learning_rate": 6.177235447089418e-06, "loss": 0.1684, "step": 9565 }, { "epoch": 2.97, "learning_rate": 6.176835367073415e-06, "loss": 0.0977, "step": 9566 }, { "epoch": 2.97, "learning_rate": 6.1764352870574115e-06, "loss": 0.0723, "step": 9567 }, { "epoch": 2.97, "learning_rate": 6.176035207041409e-06, "loss": 0.1124, "step": 9568 }, { "epoch": 2.97, "learning_rate": 6.175635127025406e-06, "loss": 0.1083, "step": 9569 }, { "epoch": 2.97, "learning_rate": 6.1752350470094024e-06, "loss": 0.0597, "step": 9570 }, { "epoch": 2.97, "learning_rate": 6.174834966993399e-06, "loss": 0.0357, "step": 9571 }, { "epoch": 2.97, "learning_rate": 6.174434886977397e-06, "loss": 0.0557, "step": 9572 }, { "epoch": 2.97, "learning_rate": 6.174034806961393e-06, "loss": 0.1169, "step": 9573 }, { "epoch": 2.97, "learning_rate": 6.17363472694539e-06, "loss": 0.0459, "step": 9574 }, { "epoch": 2.97, "learning_rate": 6.173234646929387e-06, "loss": 0.0798, "step": 9575 }, { "epoch": 2.97, "learning_rate": 6.1728345669133826e-06, "loss": 0.0937, "step": 9576 }, { "epoch": 2.97, "learning_rate": 6.172434486897379e-06, "loss": 0.0776, "step": 9577 }, { "epoch": 2.97, "learning_rate": 6.172034406881377e-06, "loss": 0.1137, "step": 9578 }, { "epoch": 2.97, "learning_rate": 6.1716343268653735e-06, "loss": 0.1265, "step": 9579 }, { "epoch": 2.97, "learning_rate": 6.17123424684937e-06, "loss": 0.0483, "step": 9580 }, { "epoch": 2.97, "learning_rate": 6.170834166833367e-06, "loss": 0.0378, "step": 9581 }, { "epoch": 2.97, "learning_rate": 6.170434086817364e-06, "loss": 0.0499, "step": 9582 }, { "epoch": 2.97, "learning_rate": 6.170034006801361e-06, "loss": 0.079, "step": 9583 }, { "epoch": 2.97, "learning_rate": 6.169633926785358e-06, "loss": 0.134, "step": 9584 }, { "epoch": 2.97, "learning_rate": 6.169233846769354e-06, "loss": 0.0724, "step": 9585 }, { "epoch": 2.98, "learning_rate": 6.168833766753352e-06, "loss": 0.0669, "step": 9586 }, { "epoch": 2.98, "learning_rate": 6.168433686737349e-06, "loss": 0.0403, "step": 9587 }, { "epoch": 2.98, "learning_rate": 6.168033606721345e-06, "loss": 0.0801, "step": 9588 }, { "epoch": 2.98, "learning_rate": 6.167633526705341e-06, "loss": 0.0551, "step": 9589 }, { "epoch": 2.98, "learning_rate": 6.167233446689338e-06, "loss": 0.1422, "step": 9590 }, { "epoch": 2.98, "learning_rate": 6.1668333666733345e-06, "loss": 0.1242, "step": 9591 }, { "epoch": 2.98, "learning_rate": 6.166433286657332e-06, "loss": 0.1227, "step": 9592 }, { "epoch": 2.98, "learning_rate": 6.166033206641329e-06, "loss": 0.1072, "step": 9593 }, { "epoch": 2.98, "learning_rate": 6.1656331266253254e-06, "loss": 0.0829, "step": 9594 }, { "epoch": 2.98, "learning_rate": 6.165233046609322e-06, "loss": 0.2064, "step": 9595 }, { "epoch": 2.98, "learning_rate": 6.16483296659332e-06, "loss": 0.0526, "step": 9596 }, { "epoch": 2.98, "learning_rate": 6.164432886577316e-06, "loss": 0.0516, "step": 9597 }, { "epoch": 2.98, "learning_rate": 6.164032806561313e-06, "loss": 0.092, "step": 9598 }, { "epoch": 2.98, "learning_rate": 6.16363272654531e-06, "loss": 0.0961, "step": 9599 }, { "epoch": 2.98, "learning_rate": 6.163232646529307e-06, "loss": 0.2828, "step": 9600 }, { "epoch": 2.98, "learning_rate": 6.162832566513302e-06, "loss": 0.0812, "step": 9601 }, { "epoch": 2.98, "learning_rate": 6.1624324864973e-06, "loss": 0.0745, "step": 9602 }, { "epoch": 2.98, "learning_rate": 6.1620324064812965e-06, "loss": 0.1315, "step": 9603 }, { "epoch": 2.98, "learning_rate": 6.161632326465293e-06, "loss": 0.0857, "step": 9604 }, { "epoch": 2.98, "learning_rate": 6.16123224644929e-06, "loss": 0.1508, "step": 9605 }, { "epoch": 2.98, "learning_rate": 6.160832166433287e-06, "loss": 0.0705, "step": 9606 }, { "epoch": 2.98, "learning_rate": 6.160432086417284e-06, "loss": 0.0979, "step": 9607 }, { "epoch": 2.98, "learning_rate": 6.160032006401281e-06, "loss": 0.0564, "step": 9608 }, { "epoch": 2.98, "learning_rate": 6.1596319263852774e-06, "loss": 0.0738, "step": 9609 }, { "epoch": 2.98, "learning_rate": 6.159231846369275e-06, "loss": 0.0388, "step": 9610 }, { "epoch": 2.98, "learning_rate": 6.158831766353272e-06, "loss": 0.142, "step": 9611 }, { "epoch": 2.98, "learning_rate": 6.158431686337268e-06, "loss": 0.0414, "step": 9612 }, { "epoch": 2.98, "learning_rate": 6.158031606321265e-06, "loss": 0.0607, "step": 9613 }, { "epoch": 2.98, "learning_rate": 6.157631526305261e-06, "loss": 0.1032, "step": 9614 }, { "epoch": 2.98, "learning_rate": 6.1572314462892576e-06, "loss": 0.0927, "step": 9615 }, { "epoch": 2.98, "learning_rate": 6.156831366273255e-06, "loss": 0.1823, "step": 9616 }, { "epoch": 2.98, "learning_rate": 6.156431286257252e-06, "loss": 0.105, "step": 9617 }, { "epoch": 2.99, "learning_rate": 6.1560312062412485e-06, "loss": 0.0747, "step": 9618 }, { "epoch": 2.99, "learning_rate": 6.155631126225245e-06, "loss": 0.059, "step": 9619 }, { "epoch": 2.99, "learning_rate": 6.155231046209243e-06, "loss": 0.0903, "step": 9620 }, { "epoch": 2.99, "learning_rate": 6.154830966193239e-06, "loss": 0.1589, "step": 9621 }, { "epoch": 2.99, "learning_rate": 6.154430886177236e-06, "loss": 0.0502, "step": 9622 }, { "epoch": 2.99, "learning_rate": 6.154030806161233e-06, "loss": 0.0575, "step": 9623 }, { "epoch": 2.99, "learning_rate": 6.15363072614523e-06, "loss": 0.1409, "step": 9624 }, { "epoch": 2.99, "learning_rate": 6.153230646129227e-06, "loss": 0.2966, "step": 9625 }, { "epoch": 2.99, "learning_rate": 6.152830566113223e-06, "loss": 0.0872, "step": 9626 }, { "epoch": 2.99, "learning_rate": 6.1524304860972195e-06, "loss": 0.1215, "step": 9627 }, { "epoch": 2.99, "learning_rate": 6.152030406081216e-06, "loss": 0.0503, "step": 9628 }, { "epoch": 2.99, "learning_rate": 6.151630326065213e-06, "loss": 0.0446, "step": 9629 }, { "epoch": 2.99, "learning_rate": 6.15123024604921e-06, "loss": 0.0897, "step": 9630 }, { "epoch": 2.99, "learning_rate": 6.150830166033207e-06, "loss": 0.0526, "step": 9631 }, { "epoch": 2.99, "learning_rate": 6.150430086017204e-06, "loss": 0.2228, "step": 9632 }, { "epoch": 2.99, "learning_rate": 6.1500300060012004e-06, "loss": 0.1204, "step": 9633 }, { "epoch": 2.99, "learning_rate": 6.149629925985198e-06, "loss": 0.0685, "step": 9634 }, { "epoch": 2.99, "learning_rate": 6.149229845969195e-06, "loss": 0.1047, "step": 9635 }, { "epoch": 2.99, "learning_rate": 6.148829765953191e-06, "loss": 0.0788, "step": 9636 }, { "epoch": 2.99, "learning_rate": 6.148429685937188e-06, "loss": 0.0887, "step": 9637 }, { "epoch": 2.99, "learning_rate": 6.1480296059211856e-06, "loss": 0.082, "step": 9638 }, { "epoch": 2.99, "learning_rate": 6.147629525905181e-06, "loss": 0.1662, "step": 9639 }, { "epoch": 2.99, "learning_rate": 6.147229445889178e-06, "loss": 0.0994, "step": 9640 }, { "epoch": 2.99, "learning_rate": 6.146829365873175e-06, "loss": 0.0631, "step": 9641 }, { "epoch": 2.99, "learning_rate": 6.1464292858571715e-06, "loss": 0.0645, "step": 9642 }, { "epoch": 2.99, "learning_rate": 6.146029205841168e-06, "loss": 0.085, "step": 9643 }, { "epoch": 2.99, "learning_rate": 6.145629125825166e-06, "loss": 0.1794, "step": 9644 }, { "epoch": 2.99, "learning_rate": 6.145229045809162e-06, "loss": 0.0707, "step": 9645 }, { "epoch": 2.99, "learning_rate": 6.144828965793159e-06, "loss": 0.0858, "step": 9646 }, { "epoch": 2.99, "learning_rate": 6.144428885777156e-06, "loss": 0.0924, "step": 9647 }, { "epoch": 2.99, "learning_rate": 6.144028805761153e-06, "loss": 0.0794, "step": 9648 }, { "epoch": 2.99, "learning_rate": 6.14362872574515e-06, "loss": 0.0934, "step": 9649 }, { "epoch": 3.0, "learning_rate": 6.143228645729147e-06, "loss": 0.0895, "step": 9650 }, { "epoch": 3.0, "learning_rate": 6.142828565713143e-06, "loss": 0.0621, "step": 9651 }, { "epoch": 3.0, "learning_rate": 6.142428485697139e-06, "loss": 0.101, "step": 9652 }, { "epoch": 3.0, "learning_rate": 6.142028405681137e-06, "loss": 0.0666, "step": 9653 }, { "epoch": 3.0, "learning_rate": 6.141628325665133e-06, "loss": 0.0538, "step": 9654 }, { "epoch": 3.0, "learning_rate": 6.14122824564913e-06, "loss": 0.103, "step": 9655 }, { "epoch": 3.0, "learning_rate": 6.140828165633127e-06, "loss": 0.1335, "step": 9656 }, { "epoch": 3.0, "learning_rate": 6.140428085617124e-06, "loss": 0.1192, "step": 9657 }, { "epoch": 3.0, "learning_rate": 6.140028005601121e-06, "loss": 0.0623, "step": 9658 }, { "epoch": 3.0, "learning_rate": 6.139627925585118e-06, "loss": 0.124, "step": 9659 }, { "epoch": 3.0, "learning_rate": 6.139227845569114e-06, "loss": 0.0544, "step": 9660 }, { "epoch": 3.0, "learning_rate": 6.138827765553111e-06, "loss": 0.0679, "step": 9661 }, { "epoch": 3.0, "learning_rate": 6.138427685537109e-06, "loss": 0.067, "step": 9662 }, { "epoch": 3.0, "learning_rate": 6.138027605521105e-06, "loss": 0.1646, "step": 9663 }, { "epoch": 3.0, "learning_rate": 6.137627525505101e-06, "loss": 0.0588, "step": 9664 }, { "epoch": 3.0, "learning_rate": 6.137227445489098e-06, "loss": 0.0912, "step": 9665 }, { "epoch": 3.0, "learning_rate": 6.1368273654730945e-06, "loss": 0.061, "step": 9666 }, { "epoch": 3.0, "learning_rate": 6.136427285457092e-06, "loss": 0.0273, "step": 9667 }, { "epoch": 3.0, "learning_rate": 6.136027205441089e-06, "loss": 0.026, "step": 9668 }, { "epoch": 3.0, "learning_rate": 6.135627125425085e-06, "loss": 0.0705, "step": 9669 }, { "epoch": 3.0, "learning_rate": 6.135227045409082e-06, "loss": 0.0428, "step": 9670 }, { "epoch": 3.0, "learning_rate": 6.13482696539308e-06, "loss": 0.0156, "step": 9671 }, { "epoch": 3.0, "learning_rate": 6.134426885377076e-06, "loss": 0.0496, "step": 9672 }, { "epoch": 3.0, "learning_rate": 6.134026805361073e-06, "loss": 0.0674, "step": 9673 }, { "epoch": 3.0, "learning_rate": 6.13362672534507e-06, "loss": 0.0829, "step": 9674 }, { "epoch": 3.0, "learning_rate": 6.133226645329066e-06, "loss": 0.0424, "step": 9675 }, { "epoch": 3.0, "learning_rate": 6.132826565313064e-06, "loss": 0.0536, "step": 9676 }, { "epoch": 3.0, "learning_rate": 6.13242648529706e-06, "loss": 0.0273, "step": 9677 }, { "epoch": 3.0, "learning_rate": 6.132026405281056e-06, "loss": 0.0352, "step": 9678 }, { "epoch": 3.0, "learning_rate": 6.131626325265053e-06, "loss": 0.0414, "step": 9679 }, { "epoch": 3.0, "learning_rate": 6.13122624524905e-06, "loss": 0.0488, "step": 9680 }, { "epoch": 3.0, "learning_rate": 6.130826165233047e-06, "loss": 0.0601, "step": 9681 }, { "epoch": 3.01, "learning_rate": 6.130426085217044e-06, "loss": 0.0242, "step": 9682 }, { "epoch": 3.01, "learning_rate": 6.130026005201041e-06, "loss": 0.0263, "step": 9683 }, { "epoch": 3.01, "learning_rate": 6.129625925185037e-06, "loss": 0.0363, "step": 9684 }, { "epoch": 3.01, "learning_rate": 6.129225845169035e-06, "loss": 0.0366, "step": 9685 }, { "epoch": 3.01, "learning_rate": 6.128825765153032e-06, "loss": 0.0303, "step": 9686 }, { "epoch": 3.01, "learning_rate": 6.128425685137028e-06, "loss": 0.0275, "step": 9687 }, { "epoch": 3.01, "learning_rate": 6.128025605121025e-06, "loss": 0.0143, "step": 9688 }, { "epoch": 3.01, "learning_rate": 6.127625525105021e-06, "loss": 0.0277, "step": 9689 }, { "epoch": 3.01, "learning_rate": 6.1272254450890175e-06, "loss": 0.0456, "step": 9690 }, { "epoch": 3.01, "learning_rate": 6.126825365073015e-06, "loss": 0.0347, "step": 9691 }, { "epoch": 3.01, "learning_rate": 6.126425285057012e-06, "loss": 0.0356, "step": 9692 }, { "epoch": 3.01, "learning_rate": 6.126025205041008e-06, "loss": 0.0964, "step": 9693 }, { "epoch": 3.01, "learning_rate": 6.125625125025005e-06, "loss": 0.0538, "step": 9694 }, { "epoch": 3.01, "learning_rate": 6.125225045009003e-06, "loss": 0.0259, "step": 9695 }, { "epoch": 3.01, "learning_rate": 6.124824964992999e-06, "loss": 0.0168, "step": 9696 }, { "epoch": 3.01, "learning_rate": 6.124424884976996e-06, "loss": 0.0245, "step": 9697 }, { "epoch": 3.01, "learning_rate": 6.124024804960993e-06, "loss": 0.04, "step": 9698 }, { "epoch": 3.01, "learning_rate": 6.12362472494499e-06, "loss": 0.0501, "step": 9699 }, { "epoch": 3.01, "learning_rate": 6.123224644928987e-06, "loss": 0.0742, "step": 9700 }, { "epoch": 3.01, "learning_rate": 6.122824564912984e-06, "loss": 0.0744, "step": 9701 }, { "epoch": 3.01, "learning_rate": 6.122424484896979e-06, "loss": 0.0308, "step": 9702 }, { "epoch": 3.01, "learning_rate": 6.122024404880976e-06, "loss": 0.1245, "step": 9703 }, { "epoch": 3.01, "learning_rate": 6.121624324864973e-06, "loss": 0.0713, "step": 9704 }, { "epoch": 3.01, "learning_rate": 6.12122424484897e-06, "loss": 0.053, "step": 9705 }, { "epoch": 3.01, "learning_rate": 6.120824164832967e-06, "loss": 0.0243, "step": 9706 }, { "epoch": 3.01, "learning_rate": 6.120424084816964e-06, "loss": 0.0314, "step": 9707 }, { "epoch": 3.01, "learning_rate": 6.12002400480096e-06, "loss": 0.0372, "step": 9708 }, { "epoch": 3.01, "learning_rate": 6.119623924784958e-06, "loss": 0.1406, "step": 9709 }, { "epoch": 3.01, "learning_rate": 6.119223844768955e-06, "loss": 0.0855, "step": 9710 }, { "epoch": 3.01, "learning_rate": 6.118823764752951e-06, "loss": 0.0307, "step": 9711 }, { "epoch": 3.01, "learning_rate": 6.118423684736948e-06, "loss": 0.0217, "step": 9712 }, { "epoch": 3.01, "learning_rate": 6.1180236047209455e-06, "loss": 0.0325, "step": 9713 }, { "epoch": 3.02, "learning_rate": 6.117623524704942e-06, "loss": 0.0318, "step": 9714 }, { "epoch": 3.02, "learning_rate": 6.117223444688938e-06, "loss": 0.0431, "step": 9715 }, { "epoch": 3.02, "learning_rate": 6.116823364672935e-06, "loss": 0.0515, "step": 9716 }, { "epoch": 3.02, "learning_rate": 6.116423284656931e-06, "loss": 0.0401, "step": 9717 }, { "epoch": 3.02, "learning_rate": 6.116023204640928e-06, "loss": 0.0372, "step": 9718 }, { "epoch": 3.02, "learning_rate": 6.115623124624926e-06, "loss": 0.0264, "step": 9719 }, { "epoch": 3.02, "learning_rate": 6.115223044608922e-06, "loss": 0.0242, "step": 9720 }, { "epoch": 3.02, "learning_rate": 6.114822964592919e-06, "loss": 0.0266, "step": 9721 }, { "epoch": 3.02, "learning_rate": 6.114422884576916e-06, "loss": 0.027, "step": 9722 }, { "epoch": 3.02, "learning_rate": 6.114022804560913e-06, "loss": 0.0738, "step": 9723 }, { "epoch": 3.02, "learning_rate": 6.11362272454491e-06, "loss": 0.0313, "step": 9724 }, { "epoch": 3.02, "learning_rate": 6.113222644528907e-06, "loss": 0.056, "step": 9725 }, { "epoch": 3.02, "learning_rate": 6.112822564512903e-06, "loss": 0.0982, "step": 9726 }, { "epoch": 3.02, "learning_rate": 6.112422484496899e-06, "loss": 0.0997, "step": 9727 }, { "epoch": 3.02, "learning_rate": 6.112022404480897e-06, "loss": 0.04, "step": 9728 }, { "epoch": 3.02, "learning_rate": 6.111622324464893e-06, "loss": 0.0152, "step": 9729 }, { "epoch": 3.02, "learning_rate": 6.11122224444889e-06, "loss": 0.0417, "step": 9730 }, { "epoch": 3.02, "learning_rate": 6.110822164432887e-06, "loss": 0.068, "step": 9731 }, { "epoch": 3.02, "learning_rate": 6.110422084416883e-06, "loss": 0.1518, "step": 9732 }, { "epoch": 3.02, "learning_rate": 6.110022004400881e-06, "loss": 0.0612, "step": 9733 }, { "epoch": 3.02, "learning_rate": 6.109621924384878e-06, "loss": 0.0109, "step": 9734 }, { "epoch": 3.02, "learning_rate": 6.109221844368874e-06, "loss": 0.0367, "step": 9735 }, { "epoch": 3.02, "learning_rate": 6.108821764352871e-06, "loss": 0.0518, "step": 9736 }, { "epoch": 3.02, "learning_rate": 6.1084216843368685e-06, "loss": 0.0262, "step": 9737 }, { "epoch": 3.02, "learning_rate": 6.108021604320865e-06, "loss": 0.0441, "step": 9738 }, { "epoch": 3.02, "learning_rate": 6.107621524304862e-06, "loss": 0.0534, "step": 9739 }, { "epoch": 3.02, "learning_rate": 6.107221444288858e-06, "loss": 0.0393, "step": 9740 }, { "epoch": 3.02, "learning_rate": 6.106821364272854e-06, "loss": 0.0324, "step": 9741 }, { "epoch": 3.02, "learning_rate": 6.106421284256852e-06, "loss": 0.0577, "step": 9742 }, { "epoch": 3.02, "learning_rate": 6.106021204240849e-06, "loss": 0.0095, "step": 9743 }, { "epoch": 3.02, "learning_rate": 6.105621124224845e-06, "loss": 0.0238, "step": 9744 }, { "epoch": 3.02, "learning_rate": 6.105221044208842e-06, "loss": 0.0316, "step": 9745 }, { "epoch": 3.02, "learning_rate": 6.104820964192839e-06, "loss": 0.011, "step": 9746 }, { "epoch": 3.03, "learning_rate": 6.104420884176836e-06, "loss": 0.0462, "step": 9747 }, { "epoch": 3.03, "learning_rate": 6.104020804160833e-06, "loss": 0.013, "step": 9748 }, { "epoch": 3.03, "learning_rate": 6.10362072414483e-06, "loss": 0.0532, "step": 9749 }, { "epoch": 3.03, "learning_rate": 6.103220644128826e-06, "loss": 0.0233, "step": 9750 }, { "epoch": 3.03, "learning_rate": 6.102820564112824e-06, "loss": 0.0193, "step": 9751 }, { "epoch": 3.03, "learning_rate": 6.10242048409682e-06, "loss": 0.0377, "step": 9752 }, { "epoch": 3.03, "learning_rate": 6.102020404080816e-06, "loss": 0.1122, "step": 9753 }, { "epoch": 3.03, "learning_rate": 6.101620324064813e-06, "loss": 0.0217, "step": 9754 }, { "epoch": 3.03, "learning_rate": 6.10122024404881e-06, "loss": 0.054, "step": 9755 }, { "epoch": 3.03, "learning_rate": 6.100820164032807e-06, "loss": 0.0322, "step": 9756 }, { "epoch": 3.03, "learning_rate": 6.100420084016804e-06, "loss": 0.0263, "step": 9757 }, { "epoch": 3.03, "learning_rate": 6.100020004000801e-06, "loss": 0.0311, "step": 9758 }, { "epoch": 3.03, "learning_rate": 6.099619923984797e-06, "loss": 0.0648, "step": 9759 }, { "epoch": 3.03, "learning_rate": 6.099219843968794e-06, "loss": 0.0462, "step": 9760 }, { "epoch": 3.03, "learning_rate": 6.0988197639527915e-06, "loss": 0.046, "step": 9761 }, { "epoch": 3.03, "learning_rate": 6.098419683936788e-06, "loss": 0.0291, "step": 9762 }, { "epoch": 3.03, "learning_rate": 6.098019603920785e-06, "loss": 0.0637, "step": 9763 }, { "epoch": 3.03, "learning_rate": 6.097619523904782e-06, "loss": 0.0509, "step": 9764 }, { "epoch": 3.03, "learning_rate": 6.0972194438887774e-06, "loss": 0.0484, "step": 9765 }, { "epoch": 3.03, "learning_rate": 6.096819363872775e-06, "loss": 0.0728, "step": 9766 }, { "epoch": 3.03, "learning_rate": 6.096419283856772e-06, "loss": 0.037, "step": 9767 }, { "epoch": 3.03, "learning_rate": 6.096019203840768e-06, "loss": 0.0457, "step": 9768 }, { "epoch": 3.03, "learning_rate": 6.095619123824765e-06, "loss": 0.0323, "step": 9769 }, { "epoch": 3.03, "learning_rate": 6.0952190438087626e-06, "loss": 0.053, "step": 9770 }, { "epoch": 3.03, "learning_rate": 6.094818963792759e-06, "loss": 0.0294, "step": 9771 }, { "epoch": 3.03, "learning_rate": 6.094418883776756e-06, "loss": 0.056, "step": 9772 }, { "epoch": 3.03, "learning_rate": 6.094018803760753e-06, "loss": 0.0646, "step": 9773 }, { "epoch": 3.03, "learning_rate": 6.093618723744749e-06, "loss": 0.0381, "step": 9774 }, { "epoch": 3.03, "learning_rate": 6.093218643728747e-06, "loss": 0.0482, "step": 9775 }, { "epoch": 3.03, "learning_rate": 6.0928185637127435e-06, "loss": 0.0269, "step": 9776 }, { "epoch": 3.03, "learning_rate": 6.09241848369674e-06, "loss": 0.0558, "step": 9777 }, { "epoch": 3.03, "learning_rate": 6.092018403680736e-06, "loss": 0.0171, "step": 9778 }, { "epoch": 3.04, "learning_rate": 6.091618323664733e-06, "loss": 0.0532, "step": 9779 }, { "epoch": 3.04, "learning_rate": 6.09121824364873e-06, "loss": 0.0416, "step": 9780 }, { "epoch": 3.04, "learning_rate": 6.090818163632727e-06, "loss": 0.0172, "step": 9781 }, { "epoch": 3.04, "learning_rate": 6.090418083616724e-06, "loss": 0.0385, "step": 9782 }, { "epoch": 3.04, "learning_rate": 6.09001800360072e-06, "loss": 0.017, "step": 9783 }, { "epoch": 3.04, "learning_rate": 6.089617923584718e-06, "loss": 0.0534, "step": 9784 }, { "epoch": 3.04, "learning_rate": 6.0892178435687145e-06, "loss": 0.0144, "step": 9785 }, { "epoch": 3.04, "learning_rate": 6.088817763552711e-06, "loss": 0.0848, "step": 9786 }, { "epoch": 3.04, "learning_rate": 6.088417683536708e-06, "loss": 0.1372, "step": 9787 }, { "epoch": 3.04, "learning_rate": 6.0880176035207055e-06, "loss": 0.025, "step": 9788 }, { "epoch": 3.04, "learning_rate": 6.087617523504702e-06, "loss": 0.0574, "step": 9789 }, { "epoch": 3.04, "learning_rate": 6.087217443488698e-06, "loss": 0.0812, "step": 9790 }, { "epoch": 3.04, "learning_rate": 6.086817363472695e-06, "loss": 0.1357, "step": 9791 }, { "epoch": 3.04, "learning_rate": 6.086417283456691e-06, "loss": 0.0825, "step": 9792 }, { "epoch": 3.04, "learning_rate": 6.086017203440688e-06, "loss": 0.0467, "step": 9793 }, { "epoch": 3.04, "learning_rate": 6.0856171234246856e-06, "loss": 0.0277, "step": 9794 }, { "epoch": 3.04, "learning_rate": 6.085217043408682e-06, "loss": 0.0848, "step": 9795 }, { "epoch": 3.04, "learning_rate": 6.084816963392679e-06, "loss": 0.0258, "step": 9796 }, { "epoch": 3.04, "learning_rate": 6.084416883376676e-06, "loss": 0.0215, "step": 9797 }, { "epoch": 3.04, "learning_rate": 6.084016803360673e-06, "loss": 0.0392, "step": 9798 }, { "epoch": 3.04, "learning_rate": 6.08361672334467e-06, "loss": 0.0268, "step": 9799 }, { "epoch": 3.04, "learning_rate": 6.0832166433286665e-06, "loss": 0.0284, "step": 9800 }, { "epoch": 3.04, "learning_rate": 6.082816563312663e-06, "loss": 0.0244, "step": 9801 }, { "epoch": 3.04, "learning_rate": 6.082416483296661e-06, "loss": 0.0681, "step": 9802 }, { "epoch": 3.04, "learning_rate": 6.082016403280656e-06, "loss": 0.0084, "step": 9803 }, { "epoch": 3.04, "learning_rate": 6.081616323264653e-06, "loss": 0.0357, "step": 9804 }, { "epoch": 3.04, "learning_rate": 6.08121624324865e-06, "loss": 0.0287, "step": 9805 }, { "epoch": 3.04, "learning_rate": 6.080816163232647e-06, "loss": 0.0178, "step": 9806 }, { "epoch": 3.04, "learning_rate": 6.080416083216643e-06, "loss": 0.0379, "step": 9807 }, { "epoch": 3.04, "learning_rate": 6.080016003200641e-06, "loss": 0.0554, "step": 9808 }, { "epoch": 3.04, "learning_rate": 6.0796159231846376e-06, "loss": 0.1938, "step": 9809 }, { "epoch": 3.04, "learning_rate": 6.079215843168634e-06, "loss": 0.0161, "step": 9810 }, { "epoch": 3.05, "learning_rate": 6.078815763152631e-06, "loss": 0.0172, "step": 9811 }, { "epoch": 3.05, "learning_rate": 6.0784156831366285e-06, "loss": 0.058, "step": 9812 }, { "epoch": 3.05, "learning_rate": 6.078015603120625e-06, "loss": 0.034, "step": 9813 }, { "epoch": 3.05, "learning_rate": 6.077615523104622e-06, "loss": 0.0557, "step": 9814 }, { "epoch": 3.05, "learning_rate": 6.0772154430886185e-06, "loss": 0.0172, "step": 9815 }, { "epoch": 3.05, "learning_rate": 6.076815363072614e-06, "loss": 0.0737, "step": 9816 }, { "epoch": 3.05, "learning_rate": 6.076415283056611e-06, "loss": 0.0289, "step": 9817 }, { "epoch": 3.05, "learning_rate": 6.076015203040609e-06, "loss": 0.0386, "step": 9818 }, { "epoch": 3.05, "learning_rate": 6.075615123024605e-06, "loss": 0.0451, "step": 9819 }, { "epoch": 3.05, "learning_rate": 6.075215043008602e-06, "loss": 0.0354, "step": 9820 }, { "epoch": 3.05, "learning_rate": 6.074814962992599e-06, "loss": 0.0538, "step": 9821 }, { "epoch": 3.05, "learning_rate": 6.074414882976596e-06, "loss": 0.0377, "step": 9822 }, { "epoch": 3.05, "learning_rate": 6.074014802960593e-06, "loss": 0.0252, "step": 9823 }, { "epoch": 3.05, "learning_rate": 6.0736147229445895e-06, "loss": 0.0519, "step": 9824 }, { "epoch": 3.05, "learning_rate": 6.073214642928586e-06, "loss": 0.0148, "step": 9825 }, { "epoch": 3.05, "learning_rate": 6.072814562912584e-06, "loss": 0.0466, "step": 9826 }, { "epoch": 3.05, "learning_rate": 6.0724144828965805e-06, "loss": 0.0285, "step": 9827 }, { "epoch": 3.05, "learning_rate": 6.072014402880576e-06, "loss": 0.0304, "step": 9828 }, { "epoch": 3.05, "learning_rate": 6.071614322864573e-06, "loss": 0.0427, "step": 9829 }, { "epoch": 3.05, "learning_rate": 6.07121424284857e-06, "loss": 0.0509, "step": 9830 }, { "epoch": 3.05, "learning_rate": 6.070814162832566e-06, "loss": 0.0477, "step": 9831 }, { "epoch": 3.05, "learning_rate": 6.070414082816564e-06, "loss": 0.0502, "step": 9832 }, { "epoch": 3.05, "learning_rate": 6.0700140028005606e-06, "loss": 0.0349, "step": 9833 }, { "epoch": 3.05, "learning_rate": 6.069613922784557e-06, "loss": 0.0254, "step": 9834 }, { "epoch": 3.05, "learning_rate": 6.069213842768554e-06, "loss": 0.032, "step": 9835 }, { "epoch": 3.05, "learning_rate": 6.0688137627525515e-06, "loss": 0.0377, "step": 9836 }, { "epoch": 3.05, "learning_rate": 6.068413682736548e-06, "loss": 0.0201, "step": 9837 }, { "epoch": 3.05, "learning_rate": 6.068013602720545e-06, "loss": 0.0327, "step": 9838 }, { "epoch": 3.05, "learning_rate": 6.0676135227045415e-06, "loss": 0.0691, "step": 9839 }, { "epoch": 3.05, "learning_rate": 6.067213442688539e-06, "loss": 0.0345, "step": 9840 }, { "epoch": 3.05, "learning_rate": 6.066813362672535e-06, "loss": 0.0334, "step": 9841 }, { "epoch": 3.05, "learning_rate": 6.066413282656532e-06, "loss": 0.0189, "step": 9842 }, { "epoch": 3.06, "learning_rate": 6.066013202640528e-06, "loss": 0.0179, "step": 9843 }, { "epoch": 3.06, "learning_rate": 6.065613122624525e-06, "loss": 0.0404, "step": 9844 }, { "epoch": 3.06, "learning_rate": 6.065213042608522e-06, "loss": 0.0279, "step": 9845 }, { "epoch": 3.06, "learning_rate": 6.064812962592519e-06, "loss": 0.0573, "step": 9846 }, { "epoch": 3.06, "learning_rate": 6.064412882576516e-06, "loss": 0.0492, "step": 9847 }, { "epoch": 3.06, "learning_rate": 6.0640128025605126e-06, "loss": 0.0675, "step": 9848 }, { "epoch": 3.06, "learning_rate": 6.063612722544509e-06, "loss": 0.032, "step": 9849 }, { "epoch": 3.06, "learning_rate": 6.063212642528507e-06, "loss": 0.0362, "step": 9850 }, { "epoch": 3.06, "learning_rate": 6.0628125625125035e-06, "loss": 0.0238, "step": 9851 }, { "epoch": 3.06, "learning_rate": 6.0624124824965e-06, "loss": 0.0417, "step": 9852 }, { "epoch": 3.06, "learning_rate": 6.062012402480496e-06, "loss": 0.0258, "step": 9853 }, { "epoch": 3.06, "learning_rate": 6.061612322464493e-06, "loss": 0.0604, "step": 9854 }, { "epoch": 3.06, "learning_rate": 6.06121224244849e-06, "loss": 0.0521, "step": 9855 }, { "epoch": 3.06, "learning_rate": 6.060812162432487e-06, "loss": 0.1596, "step": 9856 }, { "epoch": 3.06, "learning_rate": 6.060412082416484e-06, "loss": 0.1216, "step": 9857 }, { "epoch": 3.06, "learning_rate": 6.06001200240048e-06, "loss": 0.0543, "step": 9858 }, { "epoch": 3.06, "learning_rate": 6.059611922384477e-06, "loss": 0.0505, "step": 9859 }, { "epoch": 3.06, "learning_rate": 6.0592118423684745e-06, "loss": 0.0216, "step": 9860 }, { "epoch": 3.06, "learning_rate": 6.058811762352471e-06, "loss": 0.0236, "step": 9861 }, { "epoch": 3.06, "learning_rate": 6.058411682336468e-06, "loss": 0.0447, "step": 9862 }, { "epoch": 3.06, "learning_rate": 6.0580116023204645e-06, "loss": 0.1862, "step": 9863 }, { "epoch": 3.06, "learning_rate": 6.057611522304462e-06, "loss": 0.0529, "step": 9864 }, { "epoch": 3.06, "learning_rate": 6.057211442288459e-06, "loss": 0.0368, "step": 9865 }, { "epoch": 3.06, "learning_rate": 6.056811362272455e-06, "loss": 0.0759, "step": 9866 }, { "epoch": 3.06, "learning_rate": 6.056411282256451e-06, "loss": 0.1344, "step": 9867 }, { "epoch": 3.06, "learning_rate": 6.056011202240448e-06, "loss": 0.0223, "step": 9868 }, { "epoch": 3.06, "learning_rate": 6.0556111222244455e-06, "loss": 0.078, "step": 9869 }, { "epoch": 3.06, "learning_rate": 6.055211042208442e-06, "loss": 0.016, "step": 9870 }, { "epoch": 3.06, "learning_rate": 6.054810962192439e-06, "loss": 0.0769, "step": 9871 }, { "epoch": 3.06, "learning_rate": 6.0544108821764356e-06, "loss": 0.0458, "step": 9872 }, { "epoch": 3.06, "learning_rate": 6.054010802160433e-06, "loss": 0.1075, "step": 9873 }, { "epoch": 3.06, "learning_rate": 6.05361072214443e-06, "loss": 0.0474, "step": 9874 }, { "epoch": 3.06, "learning_rate": 6.0532106421284265e-06, "loss": 0.034, "step": 9875 }, { "epoch": 3.07, "learning_rate": 6.052810562112423e-06, "loss": 0.0152, "step": 9876 }, { "epoch": 3.07, "learning_rate": 6.05241048209642e-06, "loss": 0.1405, "step": 9877 }, { "epoch": 3.07, "learning_rate": 6.052010402080417e-06, "loss": 0.0295, "step": 9878 }, { "epoch": 3.07, "learning_rate": 6.051610322064413e-06, "loss": 0.0239, "step": 9879 }, { "epoch": 3.07, "learning_rate": 6.05121024204841e-06, "loss": 0.0213, "step": 9880 }, { "epoch": 3.07, "learning_rate": 6.050810162032407e-06, "loss": 0.0199, "step": 9881 }, { "epoch": 3.07, "learning_rate": 6.050410082016403e-06, "loss": 0.0333, "step": 9882 }, { "epoch": 3.07, "learning_rate": 6.050010002000401e-06, "loss": 0.0311, "step": 9883 }, { "epoch": 3.07, "learning_rate": 6.0496099219843975e-06, "loss": 0.0558, "step": 9884 }, { "epoch": 3.07, "learning_rate": 6.049209841968394e-06, "loss": 0.0489, "step": 9885 }, { "epoch": 3.07, "learning_rate": 6.048809761952391e-06, "loss": 0.0462, "step": 9886 }, { "epoch": 3.07, "learning_rate": 6.048409681936388e-06, "loss": 0.0514, "step": 9887 }, { "epoch": 3.07, "learning_rate": 6.048009601920385e-06, "loss": 0.0123, "step": 9888 }, { "epoch": 3.07, "learning_rate": 6.047609521904382e-06, "loss": 0.0278, "step": 9889 }, { "epoch": 3.07, "learning_rate": 6.0472094418883785e-06, "loss": 0.0216, "step": 9890 }, { "epoch": 3.07, "learning_rate": 6.046809361872374e-06, "loss": 0.0295, "step": 9891 }, { "epoch": 3.07, "learning_rate": 6.046409281856371e-06, "loss": 0.0337, "step": 9892 }, { "epoch": 3.07, "learning_rate": 6.0460092018403685e-06, "loss": 0.0432, "step": 9893 }, { "epoch": 3.07, "learning_rate": 6.045609121824365e-06, "loss": 0.1604, "step": 9894 }, { "epoch": 3.07, "learning_rate": 6.045209041808362e-06, "loss": 0.0137, "step": 9895 }, { "epoch": 3.07, "learning_rate": 6.044808961792359e-06, "loss": 0.0438, "step": 9896 }, { "epoch": 3.07, "learning_rate": 6.044408881776356e-06, "loss": 0.0836, "step": 9897 }, { "epoch": 3.07, "learning_rate": 6.044008801760353e-06, "loss": 0.0565, "step": 9898 }, { "epoch": 3.07, "learning_rate": 6.0436087217443495e-06, "loss": 0.0391, "step": 9899 }, { "epoch": 3.07, "learning_rate": 6.043208641728346e-06, "loss": 0.0594, "step": 9900 }, { "epoch": 3.07, "learning_rate": 6.042808561712344e-06, "loss": 0.0395, "step": 9901 }, { "epoch": 3.07, "learning_rate": 6.04240848169634e-06, "loss": 0.0369, "step": 9902 }, { "epoch": 3.07, "learning_rate": 6.042008401680337e-06, "loss": 0.0603, "step": 9903 }, { "epoch": 3.07, "learning_rate": 6.041608321664333e-06, "loss": 0.0423, "step": 9904 }, { "epoch": 3.07, "learning_rate": 6.04120824164833e-06, "loss": 0.0347, "step": 9905 }, { "epoch": 3.07, "learning_rate": 6.040808161632326e-06, "loss": 0.053, "step": 9906 }, { "epoch": 3.07, "learning_rate": 6.040408081616324e-06, "loss": 0.0288, "step": 9907 }, { "epoch": 3.08, "learning_rate": 6.0400080016003205e-06, "loss": 0.1254, "step": 9908 }, { "epoch": 3.08, "learning_rate": 6.039607921584317e-06, "loss": 0.0626, "step": 9909 }, { "epoch": 3.08, "learning_rate": 6.039207841568314e-06, "loss": 0.0227, "step": 9910 }, { "epoch": 3.08, "learning_rate": 6.038807761552311e-06, "loss": 0.021, "step": 9911 }, { "epoch": 3.08, "learning_rate": 6.038407681536308e-06, "loss": 0.0237, "step": 9912 }, { "epoch": 3.08, "learning_rate": 6.038007601520305e-06, "loss": 0.0694, "step": 9913 }, { "epoch": 3.08, "learning_rate": 6.0376075215043015e-06, "loss": 0.0209, "step": 9914 }, { "epoch": 3.08, "learning_rate": 6.037207441488299e-06, "loss": 0.0579, "step": 9915 }, { "epoch": 3.08, "learning_rate": 6.036807361472294e-06, "loss": 0.0506, "step": 9916 }, { "epoch": 3.08, "learning_rate": 6.0364072814562915e-06, "loss": 0.0376, "step": 9917 }, { "epoch": 3.08, "learning_rate": 6.036007201440288e-06, "loss": 0.0368, "step": 9918 }, { "epoch": 3.08, "learning_rate": 6.035607121424285e-06, "loss": 0.0292, "step": 9919 }, { "epoch": 3.08, "learning_rate": 6.035207041408282e-06, "loss": 0.0335, "step": 9920 }, { "epoch": 3.08, "learning_rate": 6.034806961392279e-06, "loss": 0.0513, "step": 9921 }, { "epoch": 3.08, "learning_rate": 6.034406881376276e-06, "loss": 0.0713, "step": 9922 }, { "epoch": 3.08, "learning_rate": 6.0340068013602725e-06, "loss": 0.0939, "step": 9923 }, { "epoch": 3.08, "learning_rate": 6.033606721344269e-06, "loss": 0.0227, "step": 9924 }, { "epoch": 3.08, "learning_rate": 6.033206641328267e-06, "loss": 0.0529, "step": 9925 }, { "epoch": 3.08, "learning_rate": 6.032806561312263e-06, "loss": 0.0113, "step": 9926 }, { "epoch": 3.08, "learning_rate": 6.03240648129626e-06, "loss": 0.047, "step": 9927 }, { "epoch": 3.08, "learning_rate": 6.032006401280257e-06, "loss": 0.0518, "step": 9928 }, { "epoch": 3.08, "learning_rate": 6.031606321264253e-06, "loss": 0.0306, "step": 9929 }, { "epoch": 3.08, "learning_rate": 6.031206241248249e-06, "loss": 0.1017, "step": 9930 }, { "epoch": 3.08, "learning_rate": 6.030806161232247e-06, "loss": 0.0582, "step": 9931 }, { "epoch": 3.08, "learning_rate": 6.0304060812162435e-06, "loss": 0.0547, "step": 9932 }, { "epoch": 3.08, "learning_rate": 6.03000600120024e-06, "loss": 0.0198, "step": 9933 }, { "epoch": 3.08, "learning_rate": 6.029605921184237e-06, "loss": 0.0372, "step": 9934 }, { "epoch": 3.08, "learning_rate": 6.0292058411682344e-06, "loss": 0.0855, "step": 9935 }, { "epoch": 3.08, "learning_rate": 6.028805761152231e-06, "loss": 0.0978, "step": 9936 }, { "epoch": 3.08, "learning_rate": 6.028405681136228e-06, "loss": 0.034, "step": 9937 }, { "epoch": 3.08, "learning_rate": 6.0280056011202245e-06, "loss": 0.0281, "step": 9938 }, { "epoch": 3.08, "learning_rate": 6.027605521104222e-06, "loss": 0.0936, "step": 9939 }, { "epoch": 3.09, "learning_rate": 6.027205441088219e-06, "loss": 0.026, "step": 9940 }, { "epoch": 3.09, "learning_rate": 6.026805361072215e-06, "loss": 0.0181, "step": 9941 }, { "epoch": 3.09, "learning_rate": 6.026405281056211e-06, "loss": 0.033, "step": 9942 }, { "epoch": 3.09, "learning_rate": 6.026005201040208e-06, "loss": 0.041, "step": 9943 }, { "epoch": 3.09, "learning_rate": 6.025605121024205e-06, "loss": 0.0222, "step": 9944 }, { "epoch": 3.09, "learning_rate": 6.025205041008202e-06, "loss": 0.0568, "step": 9945 }, { "epoch": 3.09, "learning_rate": 6.024804960992199e-06, "loss": 0.022, "step": 9946 }, { "epoch": 3.09, "learning_rate": 6.0244048809761955e-06, "loss": 0.0293, "step": 9947 }, { "epoch": 3.09, "learning_rate": 6.024004800960192e-06, "loss": 0.0491, "step": 9948 }, { "epoch": 3.09, "learning_rate": 6.02360472094419e-06, "loss": 0.0416, "step": 9949 }, { "epoch": 3.09, "learning_rate": 6.023204640928186e-06, "loss": 0.0223, "step": 9950 }, { "epoch": 3.09, "learning_rate": 6.022804560912183e-06, "loss": 0.0545, "step": 9951 }, { "epoch": 3.09, "learning_rate": 6.02240448089618e-06, "loss": 0.0316, "step": 9952 }, { "epoch": 3.09, "learning_rate": 6.022004400880177e-06, "loss": 0.0457, "step": 9953 }, { "epoch": 3.09, "learning_rate": 6.021604320864173e-06, "loss": 0.0473, "step": 9954 }, { "epoch": 3.09, "learning_rate": 6.02120424084817e-06, "loss": 0.0304, "step": 9955 }, { "epoch": 3.09, "learning_rate": 6.0208041608321665e-06, "loss": 0.0235, "step": 9956 }, { "epoch": 3.09, "learning_rate": 6.020404080816163e-06, "loss": 0.064, "step": 9957 }, { "epoch": 3.09, "learning_rate": 6.020004000800161e-06, "loss": 0.0426, "step": 9958 }, { "epoch": 3.09, "learning_rate": 6.0196039207841574e-06, "loss": 0.0349, "step": 9959 }, { "epoch": 3.09, "learning_rate": 6.019203840768154e-06, "loss": 0.0268, "step": 9960 }, { "epoch": 3.09, "learning_rate": 6.018803760752151e-06, "loss": 0.0858, "step": 9961 }, { "epoch": 3.09, "learning_rate": 6.0184036807361475e-06, "loss": 0.0484, "step": 9962 }, { "epoch": 3.09, "learning_rate": 6.018003600720145e-06, "loss": 0.0501, "step": 9963 }, { "epoch": 3.09, "learning_rate": 6.017603520704142e-06, "loss": 0.0672, "step": 9964 }, { "epoch": 3.09, "learning_rate": 6.017203440688138e-06, "loss": 0.023, "step": 9965 }, { "epoch": 3.09, "learning_rate": 6.016803360672135e-06, "loss": 0.0183, "step": 9966 }, { "epoch": 3.09, "learning_rate": 6.016403280656131e-06, "loss": 0.0315, "step": 9967 }, { "epoch": 3.09, "learning_rate": 6.0160032006401285e-06, "loss": 0.0111, "step": 9968 }, { "epoch": 3.09, "learning_rate": 6.015603120624125e-06, "loss": 0.0322, "step": 9969 }, { "epoch": 3.09, "learning_rate": 6.015203040608122e-06, "loss": 0.0557, "step": 9970 }, { "epoch": 3.09, "learning_rate": 6.0148029605921185e-06, "loss": 0.035, "step": 9971 }, { "epoch": 3.1, "learning_rate": 6.014402880576116e-06, "loss": 0.0379, "step": 9972 }, { "epoch": 3.1, "learning_rate": 6.014002800560113e-06, "loss": 0.0526, "step": 9973 }, { "epoch": 3.1, "learning_rate": 6.0136027205441094e-06, "loss": 0.0266, "step": 9974 }, { "epoch": 3.1, "learning_rate": 6.013202640528106e-06, "loss": 0.0267, "step": 9975 }, { "epoch": 3.1, "learning_rate": 6.012802560512103e-06, "loss": 0.0124, "step": 9976 }, { "epoch": 3.1, "learning_rate": 6.0124024804961e-06, "loss": 0.1686, "step": 9977 }, { "epoch": 3.1, "learning_rate": 6.012002400480097e-06, "loss": 0.0926, "step": 9978 }, { "epoch": 3.1, "learning_rate": 6.011602320464093e-06, "loss": 0.0384, "step": 9979 }, { "epoch": 3.1, "learning_rate": 6.0112022404480895e-06, "loss": 0.0363, "step": 9980 }, { "epoch": 3.1, "learning_rate": 6.010802160432086e-06, "loss": 0.0329, "step": 9981 }, { "epoch": 3.1, "learning_rate": 6.010402080416084e-06, "loss": 0.0229, "step": 9982 }, { "epoch": 3.1, "learning_rate": 6.0100020004000805e-06, "loss": 0.0685, "step": 9983 }, { "epoch": 3.1, "learning_rate": 6.009601920384077e-06, "loss": 0.0202, "step": 9984 }, { "epoch": 3.1, "learning_rate": 6.009201840368074e-06, "loss": 0.0313, "step": 9985 }, { "epoch": 3.1, "learning_rate": 6.008801760352071e-06, "loss": 0.0365, "step": 9986 }, { "epoch": 3.1, "learning_rate": 6.008401680336068e-06, "loss": 0.0307, "step": 9987 }, { "epoch": 3.1, "learning_rate": 6.008001600320065e-06, "loss": 0.0201, "step": 9988 }, { "epoch": 3.1, "learning_rate": 6.007601520304061e-06, "loss": 0.051, "step": 9989 }, { "epoch": 3.1, "learning_rate": 6.007201440288058e-06, "loss": 0.0236, "step": 9990 }, { "epoch": 3.1, "learning_rate": 6.006801360272056e-06, "loss": 0.0311, "step": 9991 }, { "epoch": 3.1, "learning_rate": 6.0064012802560515e-06, "loss": 0.0416, "step": 9992 }, { "epoch": 3.1, "learning_rate": 6.006001200240048e-06, "loss": 0.0125, "step": 9993 }, { "epoch": 3.1, "learning_rate": 6.005601120224045e-06, "loss": 0.0576, "step": 9994 }, { "epoch": 3.1, "learning_rate": 6.0052010402080415e-06, "loss": 0.0333, "step": 9995 }, { "epoch": 3.1, "learning_rate": 6.004800960192039e-06, "loss": 0.0482, "step": 9996 }, { "epoch": 3.1, "learning_rate": 6.004400880176036e-06, "loss": 0.05, "step": 9997 }, { "epoch": 3.1, "learning_rate": 6.0040008001600324e-06, "loss": 0.0307, "step": 9998 }, { "epoch": 3.1, "learning_rate": 6.003600720144029e-06, "loss": 0.0462, "step": 9999 }, { "epoch": 3.1, "learning_rate": 6.003200640128027e-06, "loss": 0.0054, "step": 10000 }, { "epoch": 3.1, "learning_rate": 6.002800560112023e-06, "loss": 0.0314, "step": 10001 }, { "epoch": 3.1, "learning_rate": 6.00240048009602e-06, "loss": 0.0179, "step": 10002 }, { "epoch": 3.1, "learning_rate": 6.002000400080017e-06, "loss": 0.0528, "step": 10003 }, { "epoch": 3.11, "learning_rate": 6.001600320064013e-06, "loss": 0.0935, "step": 10004 }, { "epoch": 3.11, "learning_rate": 6.001200240048009e-06, "loss": 0.0284, "step": 10005 }, { "epoch": 3.11, "learning_rate": 6.000800160032007e-06, "loss": 0.0319, "step": 10006 }, { "epoch": 3.11, "learning_rate": 6.0004000800160035e-06, "loss": 0.0354, "step": 10007 }, { "epoch": 3.11, "learning_rate": 6e-06, "loss": 0.0179, "step": 10008 }, { "epoch": 3.11, "learning_rate": 5.999599919983997e-06, "loss": 0.0479, "step": 10009 }, { "epoch": 3.11, "learning_rate": 5.999199839967994e-06, "loss": 0.02, "step": 10010 }, { "epoch": 3.11, "learning_rate": 5.998799759951991e-06, "loss": 0.0297, "step": 10011 }, { "epoch": 3.11, "learning_rate": 5.998399679935988e-06, "loss": 0.0286, "step": 10012 }, { "epoch": 3.11, "learning_rate": 5.9979995999199844e-06, "loss": 0.0284, "step": 10013 }, { "epoch": 3.11, "learning_rate": 5.997599519903982e-06, "loss": 0.0288, "step": 10014 }, { "epoch": 3.11, "learning_rate": 5.997199439887979e-06, "loss": 0.023, "step": 10015 }, { "epoch": 3.11, "learning_rate": 5.996799359871975e-06, "loss": 0.0322, "step": 10016 }, { "epoch": 3.11, "learning_rate": 5.996399279855971e-06, "loss": 0.044, "step": 10017 }, { "epoch": 3.11, "learning_rate": 5.995999199839968e-06, "loss": 0.0341, "step": 10018 }, { "epoch": 3.11, "learning_rate": 5.9955991198239645e-06, "loss": 0.0407, "step": 10019 }, { "epoch": 3.11, "learning_rate": 5.995199039807962e-06, "loss": 0.0379, "step": 10020 }, { "epoch": 3.11, "learning_rate": 5.994798959791959e-06, "loss": 0.1385, "step": 10021 }, { "epoch": 3.11, "learning_rate": 5.9943988797759555e-06, "loss": 0.0897, "step": 10022 }, { "epoch": 3.11, "learning_rate": 5.993998799759952e-06, "loss": 0.0548, "step": 10023 }, { "epoch": 3.11, "learning_rate": 5.99359871974395e-06, "loss": 0.0274, "step": 10024 }, { "epoch": 3.11, "learning_rate": 5.993198639727946e-06, "loss": 0.0372, "step": 10025 }, { "epoch": 3.11, "learning_rate": 5.992798559711943e-06, "loss": 0.0426, "step": 10026 }, { "epoch": 3.11, "learning_rate": 5.99239847969594e-06, "loss": 0.1283, "step": 10027 }, { "epoch": 3.11, "learning_rate": 5.991998399679937e-06, "loss": 0.0489, "step": 10028 }, { "epoch": 3.11, "learning_rate": 5.991598319663934e-06, "loss": 0.0503, "step": 10029 }, { "epoch": 3.11, "learning_rate": 5.99119823964793e-06, "loss": 0.0221, "step": 10030 }, { "epoch": 3.11, "learning_rate": 5.9907981596319265e-06, "loss": 0.0437, "step": 10031 }, { "epoch": 3.11, "learning_rate": 5.990398079615923e-06, "loss": 0.059, "step": 10032 }, { "epoch": 3.11, "learning_rate": 5.98999799959992e-06, "loss": 0.0342, "step": 10033 }, { "epoch": 3.11, "learning_rate": 5.989597919583917e-06, "loss": 0.0572, "step": 10034 }, { "epoch": 3.11, "learning_rate": 5.989197839567914e-06, "loss": 0.0214, "step": 10035 }, { "epoch": 3.11, "learning_rate": 5.988797759551911e-06, "loss": 0.0271, "step": 10036 }, { "epoch": 3.12, "learning_rate": 5.9883976795359074e-06, "loss": 0.047, "step": 10037 }, { "epoch": 3.12, "learning_rate": 5.987997599519905e-06, "loss": 0.1449, "step": 10038 }, { "epoch": 3.12, "learning_rate": 5.987597519503902e-06, "loss": 0.0428, "step": 10039 }, { "epoch": 3.12, "learning_rate": 5.987197439487898e-06, "loss": 0.0517, "step": 10040 }, { "epoch": 3.12, "learning_rate": 5.986797359471895e-06, "loss": 0.0432, "step": 10041 }, { "epoch": 3.12, "learning_rate": 5.9863972794558926e-06, "loss": 0.0308, "step": 10042 }, { "epoch": 3.12, "learning_rate": 5.985997199439888e-06, "loss": 0.0879, "step": 10043 }, { "epoch": 3.12, "learning_rate": 5.985597119423885e-06, "loss": 0.1578, "step": 10044 }, { "epoch": 3.12, "learning_rate": 5.985197039407882e-06, "loss": 0.0264, "step": 10045 }, { "epoch": 3.12, "learning_rate": 5.9847969593918785e-06, "loss": 0.0308, "step": 10046 }, { "epoch": 3.12, "learning_rate": 5.984396879375875e-06, "loss": 0.0259, "step": 10047 }, { "epoch": 3.12, "learning_rate": 5.983996799359873e-06, "loss": 0.057, "step": 10048 }, { "epoch": 3.12, "learning_rate": 5.983596719343869e-06, "loss": 0.0394, "step": 10049 }, { "epoch": 3.12, "learning_rate": 5.983196639327866e-06, "loss": 0.0383, "step": 10050 }, { "epoch": 3.12, "learning_rate": 5.982796559311863e-06, "loss": 0.0598, "step": 10051 }, { "epoch": 3.12, "learning_rate": 5.98239647929586e-06, "loss": 0.0206, "step": 10052 }, { "epoch": 3.12, "learning_rate": 5.981996399279857e-06, "loss": 0.0673, "step": 10053 }, { "epoch": 3.12, "learning_rate": 5.981596319263854e-06, "loss": 0.0325, "step": 10054 }, { "epoch": 3.12, "learning_rate": 5.9811962392478495e-06, "loss": 0.0316, "step": 10055 }, { "epoch": 3.12, "learning_rate": 5.980796159231846e-06, "loss": 0.0391, "step": 10056 }, { "epoch": 3.12, "learning_rate": 5.980396079215844e-06, "loss": 0.0307, "step": 10057 }, { "epoch": 3.12, "learning_rate": 5.97999599919984e-06, "loss": 0.0363, "step": 10058 }, { "epoch": 3.12, "learning_rate": 5.979595919183837e-06, "loss": 0.0825, "step": 10059 }, { "epoch": 3.12, "learning_rate": 5.979195839167834e-06, "loss": 0.0443, "step": 10060 }, { "epoch": 3.12, "learning_rate": 5.9787957591518305e-06, "loss": 0.1091, "step": 10061 }, { "epoch": 3.12, "learning_rate": 5.978395679135828e-06, "loss": 0.0645, "step": 10062 }, { "epoch": 3.12, "learning_rate": 5.977995599119825e-06, "loss": 0.0938, "step": 10063 }, { "epoch": 3.12, "learning_rate": 5.977595519103821e-06, "loss": 0.0254, "step": 10064 }, { "epoch": 3.12, "learning_rate": 5.977195439087818e-06, "loss": 0.0775, "step": 10065 }, { "epoch": 3.12, "learning_rate": 5.976795359071816e-06, "loss": 0.0226, "step": 10066 }, { "epoch": 3.12, "learning_rate": 5.976395279055812e-06, "loss": 0.024, "step": 10067 }, { "epoch": 3.12, "learning_rate": 5.975995199039808e-06, "loss": 0.0621, "step": 10068 }, { "epoch": 3.13, "learning_rate": 5.975595119023805e-06, "loss": 0.0573, "step": 10069 }, { "epoch": 3.13, "learning_rate": 5.9751950390078015e-06, "loss": 0.0293, "step": 10070 }, { "epoch": 3.13, "learning_rate": 5.974794958991799e-06, "loss": 0.0079, "step": 10071 }, { "epoch": 3.13, "learning_rate": 5.974394878975796e-06, "loss": 0.0237, "step": 10072 }, { "epoch": 3.13, "learning_rate": 5.973994798959792e-06, "loss": 0.0303, "step": 10073 }, { "epoch": 3.13, "learning_rate": 5.973594718943789e-06, "loss": 0.0229, "step": 10074 }, { "epoch": 3.13, "learning_rate": 5.973194638927786e-06, "loss": 0.0351, "step": 10075 }, { "epoch": 3.13, "learning_rate": 5.972794558911783e-06, "loss": 0.0214, "step": 10076 }, { "epoch": 3.13, "learning_rate": 5.97239447889578e-06, "loss": 0.0313, "step": 10077 }, { "epoch": 3.13, "learning_rate": 5.971994398879777e-06, "loss": 0.0723, "step": 10078 }, { "epoch": 3.13, "learning_rate": 5.971594318863773e-06, "loss": 0.067, "step": 10079 }, { "epoch": 3.13, "learning_rate": 5.971194238847769e-06, "loss": 0.0643, "step": 10080 }, { "epoch": 3.13, "learning_rate": 5.970794158831767e-06, "loss": 0.045, "step": 10081 }, { "epoch": 3.13, "learning_rate": 5.970394078815763e-06, "loss": 0.0601, "step": 10082 }, { "epoch": 3.13, "learning_rate": 5.96999399879976e-06, "loss": 0.0819, "step": 10083 }, { "epoch": 3.13, "learning_rate": 5.969593918783757e-06, "loss": 0.0386, "step": 10084 }, { "epoch": 3.13, "learning_rate": 5.969193838767754e-06, "loss": 0.0341, "step": 10085 }, { "epoch": 3.13, "learning_rate": 5.968793758751751e-06, "loss": 0.061, "step": 10086 }, { "epoch": 3.13, "learning_rate": 5.968393678735748e-06, "loss": 0.0171, "step": 10087 }, { "epoch": 3.13, "learning_rate": 5.967993598719744e-06, "loss": 0.031, "step": 10088 }, { "epoch": 3.13, "learning_rate": 5.967593518703742e-06, "loss": 0.0357, "step": 10089 }, { "epoch": 3.13, "learning_rate": 5.967193438687739e-06, "loss": 0.0653, "step": 10090 }, { "epoch": 3.13, "learning_rate": 5.966793358671735e-06, "loss": 0.0287, "step": 10091 }, { "epoch": 3.13, "learning_rate": 5.966393278655732e-06, "loss": 0.2189, "step": 10092 }, { "epoch": 3.13, "learning_rate": 5.965993198639728e-06, "loss": 0.0305, "step": 10093 }, { "epoch": 3.13, "learning_rate": 5.9655931186237245e-06, "loss": 0.0304, "step": 10094 }, { "epoch": 3.13, "learning_rate": 5.965193038607722e-06, "loss": 0.0293, "step": 10095 }, { "epoch": 3.13, "learning_rate": 5.964792958591719e-06, "loss": 0.0445, "step": 10096 }, { "epoch": 3.13, "learning_rate": 5.964392878575715e-06, "loss": 0.0475, "step": 10097 }, { "epoch": 3.13, "learning_rate": 5.963992798559712e-06, "loss": 0.048, "step": 10098 }, { "epoch": 3.13, "learning_rate": 5.96359271854371e-06, "loss": 0.0347, "step": 10099 }, { "epoch": 3.13, "learning_rate": 5.963192638527706e-06, "loss": 0.0175, "step": 10100 }, { "epoch": 3.14, "learning_rate": 5.962792558511703e-06, "loss": 0.0521, "step": 10101 }, { "epoch": 3.14, "learning_rate": 5.9623924784957e-06, "loss": 0.0368, "step": 10102 }, { "epoch": 3.14, "learning_rate": 5.961992398479697e-06, "loss": 0.0462, "step": 10103 }, { "epoch": 3.14, "learning_rate": 5.961592318463694e-06, "loss": 0.0119, "step": 10104 }, { "epoch": 3.14, "learning_rate": 5.961192238447691e-06, "loss": 0.0065, "step": 10105 }, { "epoch": 3.14, "learning_rate": 5.960792158431686e-06, "loss": 0.0846, "step": 10106 }, { "epoch": 3.14, "learning_rate": 5.960392078415683e-06, "loss": 0.114, "step": 10107 }, { "epoch": 3.14, "learning_rate": 5.95999199839968e-06, "loss": 0.0134, "step": 10108 }, { "epoch": 3.14, "learning_rate": 5.959591918383677e-06, "loss": 0.0391, "step": 10109 }, { "epoch": 3.14, "learning_rate": 5.959191838367674e-06, "loss": 0.0383, "step": 10110 }, { "epoch": 3.14, "learning_rate": 5.958791758351671e-06, "loss": 0.0387, "step": 10111 }, { "epoch": 3.14, "learning_rate": 5.958391678335667e-06, "loss": 0.0127, "step": 10112 }, { "epoch": 3.14, "learning_rate": 5.957991598319665e-06, "loss": 0.0695, "step": 10113 }, { "epoch": 3.14, "learning_rate": 5.957591518303662e-06, "loss": 0.0337, "step": 10114 }, { "epoch": 3.14, "learning_rate": 5.957191438287658e-06, "loss": 0.0415, "step": 10115 }, { "epoch": 3.14, "learning_rate": 5.956791358271655e-06, "loss": 0.0717, "step": 10116 }, { "epoch": 3.14, "learning_rate": 5.9563912782556525e-06, "loss": 0.1395, "step": 10117 }, { "epoch": 3.14, "learning_rate": 5.9559911982396475e-06, "loss": 0.0589, "step": 10118 }, { "epoch": 3.14, "learning_rate": 5.955591118223645e-06, "loss": 0.0742, "step": 10119 }, { "epoch": 3.14, "learning_rate": 5.955191038207642e-06, "loss": 0.0697, "step": 10120 }, { "epoch": 3.14, "learning_rate": 5.954790958191638e-06, "loss": 0.0241, "step": 10121 }, { "epoch": 3.14, "learning_rate": 5.954390878175635e-06, "loss": 0.1766, "step": 10122 }, { "epoch": 3.14, "learning_rate": 5.953990798159633e-06, "loss": 0.0356, "step": 10123 }, { "epoch": 3.14, "learning_rate": 5.953590718143629e-06, "loss": 0.0175, "step": 10124 }, { "epoch": 3.14, "learning_rate": 5.953190638127626e-06, "loss": 0.0577, "step": 10125 }, { "epoch": 3.14, "learning_rate": 5.952790558111623e-06, "loss": 0.0744, "step": 10126 }, { "epoch": 3.14, "learning_rate": 5.95239047809562e-06, "loss": 0.0665, "step": 10127 }, { "epoch": 3.14, "learning_rate": 5.951990398079617e-06, "loss": 0.0444, "step": 10128 }, { "epoch": 3.14, "learning_rate": 5.951590318063614e-06, "loss": 0.0386, "step": 10129 }, { "epoch": 3.14, "learning_rate": 5.95119023804761e-06, "loss": 0.0285, "step": 10130 }, { "epoch": 3.14, "learning_rate": 5.950790158031606e-06, "loss": 0.0887, "step": 10131 }, { "epoch": 3.14, "learning_rate": 5.950390078015603e-06, "loss": 0.0132, "step": 10132 }, { "epoch": 3.15, "learning_rate": 5.9499899979996e-06, "loss": 0.0426, "step": 10133 }, { "epoch": 3.15, "learning_rate": 5.949589917983597e-06, "loss": 0.0637, "step": 10134 }, { "epoch": 3.15, "learning_rate": 5.949189837967594e-06, "loss": 0.018, "step": 10135 }, { "epoch": 3.15, "learning_rate": 5.94878975795159e-06, "loss": 0.0469, "step": 10136 }, { "epoch": 3.15, "learning_rate": 5.948389677935588e-06, "loss": 0.0234, "step": 10137 }, { "epoch": 3.15, "learning_rate": 5.947989597919585e-06, "loss": 0.1408, "step": 10138 }, { "epoch": 3.15, "learning_rate": 5.947589517903581e-06, "loss": 0.0558, "step": 10139 }, { "epoch": 3.15, "learning_rate": 5.947189437887578e-06, "loss": 0.0619, "step": 10140 }, { "epoch": 3.15, "learning_rate": 5.9467893578715755e-06, "loss": 0.0449, "step": 10141 }, { "epoch": 3.15, "learning_rate": 5.946389277855572e-06, "loss": 0.0677, "step": 10142 }, { "epoch": 3.15, "learning_rate": 5.945989197839568e-06, "loss": 0.0322, "step": 10143 }, { "epoch": 3.15, "learning_rate": 5.945589117823565e-06, "loss": 0.0654, "step": 10144 }, { "epoch": 3.15, "learning_rate": 5.945189037807561e-06, "loss": 0.0264, "step": 10145 }, { "epoch": 3.15, "learning_rate": 5.944788957791558e-06, "loss": 0.0407, "step": 10146 }, { "epoch": 3.15, "learning_rate": 5.944388877775556e-06, "loss": 0.0664, "step": 10147 }, { "epoch": 3.15, "learning_rate": 5.943988797759552e-06, "loss": 0.0256, "step": 10148 }, { "epoch": 3.15, "learning_rate": 5.943588717743549e-06, "loss": 0.0549, "step": 10149 }, { "epoch": 3.15, "learning_rate": 5.943188637727546e-06, "loss": 0.0696, "step": 10150 }, { "epoch": 3.15, "learning_rate": 5.942788557711543e-06, "loss": 0.0454, "step": 10151 }, { "epoch": 3.15, "learning_rate": 5.94238847769554e-06, "loss": 0.022, "step": 10152 }, { "epoch": 3.15, "learning_rate": 5.941988397679537e-06, "loss": 0.0352, "step": 10153 }, { "epoch": 3.15, "learning_rate": 5.941588317663533e-06, "loss": 0.0165, "step": 10154 }, { "epoch": 3.15, "learning_rate": 5.941188237647531e-06, "loss": 0.1696, "step": 10155 }, { "epoch": 3.15, "learning_rate": 5.940788157631527e-06, "loss": 0.0684, "step": 10156 }, { "epoch": 3.15, "learning_rate": 5.940388077615523e-06, "loss": 0.0488, "step": 10157 }, { "epoch": 3.15, "learning_rate": 5.93998799759952e-06, "loss": 0.0444, "step": 10158 }, { "epoch": 3.15, "learning_rate": 5.939587917583517e-06, "loss": 0.034, "step": 10159 }, { "epoch": 3.15, "learning_rate": 5.939187837567513e-06, "loss": 0.0422, "step": 10160 }, { "epoch": 3.15, "learning_rate": 5.938787757551511e-06, "loss": 0.0443, "step": 10161 }, { "epoch": 3.15, "learning_rate": 5.938387677535508e-06, "loss": 0.0612, "step": 10162 }, { "epoch": 3.15, "learning_rate": 5.937987597519504e-06, "loss": 0.0529, "step": 10163 }, { "epoch": 3.15, "learning_rate": 5.937587517503501e-06, "loss": 0.0757, "step": 10164 }, { "epoch": 3.15, "learning_rate": 5.9371874374874985e-06, "loss": 0.0278, "step": 10165 }, { "epoch": 3.16, "learning_rate": 5.936787357471495e-06, "loss": 0.1541, "step": 10166 }, { "epoch": 3.16, "learning_rate": 5.936387277455492e-06, "loss": 0.0801, "step": 10167 }, { "epoch": 3.16, "learning_rate": 5.935987197439489e-06, "loss": 0.0315, "step": 10168 }, { "epoch": 3.16, "learning_rate": 5.9355871174234844e-06, "loss": 0.0175, "step": 10169 }, { "epoch": 3.16, "learning_rate": 5.935187037407482e-06, "loss": 0.0366, "step": 10170 }, { "epoch": 3.16, "learning_rate": 5.934786957391479e-06, "loss": 0.0528, "step": 10171 }, { "epoch": 3.16, "learning_rate": 5.934386877375475e-06, "loss": 0.0292, "step": 10172 }, { "epoch": 3.16, "learning_rate": 5.933986797359472e-06, "loss": 0.0448, "step": 10173 }, { "epoch": 3.16, "learning_rate": 5.9335867173434696e-06, "loss": 0.0344, "step": 10174 }, { "epoch": 3.16, "learning_rate": 5.933186637327466e-06, "loss": 0.0189, "step": 10175 }, { "epoch": 3.16, "learning_rate": 5.932786557311463e-06, "loss": 0.0377, "step": 10176 }, { "epoch": 3.16, "learning_rate": 5.93238647729546e-06, "loss": 0.0425, "step": 10177 }, { "epoch": 3.16, "learning_rate": 5.931986397279456e-06, "loss": 0.0193, "step": 10178 }, { "epoch": 3.16, "learning_rate": 5.931586317263454e-06, "loss": 0.0163, "step": 10179 }, { "epoch": 3.16, "learning_rate": 5.9311862372474505e-06, "loss": 0.0277, "step": 10180 }, { "epoch": 3.16, "learning_rate": 5.930786157231446e-06, "loss": 0.0459, "step": 10181 }, { "epoch": 3.16, "learning_rate": 5.930386077215443e-06, "loss": 0.0245, "step": 10182 }, { "epoch": 3.16, "learning_rate": 5.92998599719944e-06, "loss": 0.0268, "step": 10183 }, { "epoch": 3.16, "learning_rate": 5.929585917183437e-06, "loss": 0.0719, "step": 10184 }, { "epoch": 3.16, "learning_rate": 5.929185837167434e-06, "loss": 0.0363, "step": 10185 }, { "epoch": 3.16, "learning_rate": 5.928785757151431e-06, "loss": 0.0241, "step": 10186 }, { "epoch": 3.16, "learning_rate": 5.928385677135427e-06, "loss": 0.0604, "step": 10187 }, { "epoch": 3.16, "learning_rate": 5.927985597119425e-06, "loss": 0.0377, "step": 10188 }, { "epoch": 3.16, "learning_rate": 5.9275855171034215e-06, "loss": 0.052, "step": 10189 }, { "epoch": 3.16, "learning_rate": 5.927185437087418e-06, "loss": 0.0279, "step": 10190 }, { "epoch": 3.16, "learning_rate": 5.926785357071415e-06, "loss": 0.0354, "step": 10191 }, { "epoch": 3.16, "learning_rate": 5.926385277055412e-06, "loss": 0.0698, "step": 10192 }, { "epoch": 3.16, "learning_rate": 5.925985197039409e-06, "loss": 0.0802, "step": 10193 }, { "epoch": 3.16, "learning_rate": 5.925585117023405e-06, "loss": 0.033, "step": 10194 }, { "epoch": 3.16, "learning_rate": 5.925185037007402e-06, "loss": 0.0748, "step": 10195 }, { "epoch": 3.16, "learning_rate": 5.924784956991398e-06, "loss": 0.0279, "step": 10196 }, { "epoch": 3.16, "learning_rate": 5.924384876975395e-06, "loss": 0.0317, "step": 10197 }, { "epoch": 3.17, "learning_rate": 5.9239847969593926e-06, "loss": 0.0665, "step": 10198 }, { "epoch": 3.17, "learning_rate": 5.923584716943389e-06, "loss": 0.02, "step": 10199 }, { "epoch": 3.17, "learning_rate": 5.923184636927386e-06, "loss": 0.0383, "step": 10200 }, { "epoch": 3.17, "learning_rate": 5.922784556911383e-06, "loss": 0.0444, "step": 10201 }, { "epoch": 3.17, "learning_rate": 5.92238447689538e-06, "loss": 0.0301, "step": 10202 }, { "epoch": 3.17, "learning_rate": 5.921984396879377e-06, "loss": 0.052, "step": 10203 }, { "epoch": 3.17, "learning_rate": 5.9215843168633735e-06, "loss": 0.0403, "step": 10204 }, { "epoch": 3.17, "learning_rate": 5.92118423684737e-06, "loss": 0.0605, "step": 10205 }, { "epoch": 3.17, "learning_rate": 5.920784156831366e-06, "loss": 0.0885, "step": 10206 }, { "epoch": 3.17, "learning_rate": 5.920384076815363e-06, "loss": 0.0464, "step": 10207 }, { "epoch": 3.17, "learning_rate": 5.91998399679936e-06, "loss": 0.0501, "step": 10208 }, { "epoch": 3.17, "learning_rate": 5.919583916783357e-06, "loss": 0.0328, "step": 10209 }, { "epoch": 3.17, "learning_rate": 5.919183836767354e-06, "loss": 0.0529, "step": 10210 }, { "epoch": 3.17, "learning_rate": 5.91878375675135e-06, "loss": 0.0247, "step": 10211 }, { "epoch": 3.17, "learning_rate": 5.918383676735348e-06, "loss": 0.0371, "step": 10212 }, { "epoch": 3.17, "learning_rate": 5.9179835967193446e-06, "loss": 0.1442, "step": 10213 }, { "epoch": 3.17, "learning_rate": 5.917583516703341e-06, "loss": 0.041, "step": 10214 }, { "epoch": 3.17, "learning_rate": 5.917183436687338e-06, "loss": 0.0466, "step": 10215 }, { "epoch": 3.17, "learning_rate": 5.9167833566713355e-06, "loss": 0.031, "step": 10216 }, { "epoch": 3.17, "learning_rate": 5.916383276655332e-06, "loss": 0.1402, "step": 10217 }, { "epoch": 3.17, "learning_rate": 5.915983196639329e-06, "loss": 0.0849, "step": 10218 }, { "epoch": 3.17, "learning_rate": 5.915583116623325e-06, "loss": 0.0322, "step": 10219 }, { "epoch": 3.17, "learning_rate": 5.915183036607321e-06, "loss": 0.0707, "step": 10220 }, { "epoch": 3.17, "learning_rate": 5.914782956591318e-06, "loss": 0.0721, "step": 10221 }, { "epoch": 3.17, "learning_rate": 5.914382876575316e-06, "loss": 0.0393, "step": 10222 }, { "epoch": 3.17, "learning_rate": 5.913982796559312e-06, "loss": 0.0664, "step": 10223 }, { "epoch": 3.17, "learning_rate": 5.913582716543309e-06, "loss": 0.0289, "step": 10224 }, { "epoch": 3.17, "learning_rate": 5.913182636527306e-06, "loss": 0.0736, "step": 10225 }, { "epoch": 3.17, "learning_rate": 5.912782556511303e-06, "loss": 0.0384, "step": 10226 }, { "epoch": 3.17, "learning_rate": 5.9123824764953e-06, "loss": 0.0298, "step": 10227 }, { "epoch": 3.17, "learning_rate": 5.9119823964792965e-06, "loss": 0.0331, "step": 10228 }, { "epoch": 3.17, "learning_rate": 5.911582316463293e-06, "loss": 0.0555, "step": 10229 }, { "epoch": 3.18, "learning_rate": 5.911182236447291e-06, "loss": 0.0116, "step": 10230 }, { "epoch": 3.18, "learning_rate": 5.9107821564312874e-06, "loss": 0.0348, "step": 10231 }, { "epoch": 3.18, "learning_rate": 5.910382076415283e-06, "loss": 0.0272, "step": 10232 }, { "epoch": 3.18, "learning_rate": 5.90998199639928e-06, "loss": 0.1129, "step": 10233 }, { "epoch": 3.18, "learning_rate": 5.909581916383277e-06, "loss": 0.0655, "step": 10234 }, { "epoch": 3.18, "learning_rate": 5.909181836367273e-06, "loss": 0.0348, "step": 10235 }, { "epoch": 3.18, "learning_rate": 5.908781756351271e-06, "loss": 0.0401, "step": 10236 }, { "epoch": 3.18, "learning_rate": 5.9083816763352676e-06, "loss": 0.0406, "step": 10237 }, { "epoch": 3.18, "learning_rate": 5.907981596319264e-06, "loss": 0.0477, "step": 10238 }, { "epoch": 3.18, "learning_rate": 5.907581516303261e-06, "loss": 0.0802, "step": 10239 }, { "epoch": 3.18, "learning_rate": 5.9071814362872585e-06, "loss": 0.0371, "step": 10240 }, { "epoch": 3.18, "learning_rate": 5.906781356271255e-06, "loss": 0.1264, "step": 10241 }, { "epoch": 3.18, "learning_rate": 5.906381276255252e-06, "loss": 0.041, "step": 10242 }, { "epoch": 3.18, "learning_rate": 5.9059811962392485e-06, "loss": 0.0408, "step": 10243 }, { "epoch": 3.18, "learning_rate": 5.905581116223244e-06, "loss": 0.0547, "step": 10244 }, { "epoch": 3.18, "learning_rate": 5.905181036207241e-06, "loss": 0.0571, "step": 10245 }, { "epoch": 3.18, "learning_rate": 5.904780956191239e-06, "loss": 0.1506, "step": 10246 }, { "epoch": 3.18, "learning_rate": 5.904380876175235e-06, "loss": 0.0385, "step": 10247 }, { "epoch": 3.18, "learning_rate": 5.903980796159232e-06, "loss": 0.048, "step": 10248 }, { "epoch": 3.18, "learning_rate": 5.903580716143229e-06, "loss": 0.0315, "step": 10249 }, { "epoch": 3.18, "learning_rate": 5.903180636127226e-06, "loss": 0.0229, "step": 10250 }, { "epoch": 3.18, "learning_rate": 5.902780556111223e-06, "loss": 0.0786, "step": 10251 }, { "epoch": 3.18, "learning_rate": 5.9023804760952196e-06, "loss": 0.0261, "step": 10252 }, { "epoch": 3.18, "learning_rate": 5.901980396079216e-06, "loss": 0.0362, "step": 10253 }, { "epoch": 3.18, "learning_rate": 5.901580316063214e-06, "loss": 0.0595, "step": 10254 }, { "epoch": 3.18, "learning_rate": 5.9011802360472105e-06, "loss": 0.0348, "step": 10255 }, { "epoch": 3.18, "learning_rate": 5.900780156031207e-06, "loss": 0.0357, "step": 10256 }, { "epoch": 3.18, "learning_rate": 5.900380076015203e-06, "loss": 0.0502, "step": 10257 }, { "epoch": 3.18, "learning_rate": 5.8999799959992e-06, "loss": 0.0427, "step": 10258 }, { "epoch": 3.18, "learning_rate": 5.899579915983197e-06, "loss": 0.0571, "step": 10259 }, { "epoch": 3.18, "learning_rate": 5.899179835967194e-06, "loss": 0.174, "step": 10260 }, { "epoch": 3.18, "learning_rate": 5.898779755951191e-06, "loss": 0.0322, "step": 10261 }, { "epoch": 3.19, "learning_rate": 5.898379675935187e-06, "loss": 0.0687, "step": 10262 }, { "epoch": 3.19, "learning_rate": 5.897979595919184e-06, "loss": 0.0244, "step": 10263 }, { "epoch": 3.19, "learning_rate": 5.8975795159031815e-06, "loss": 0.1785, "step": 10264 }, { "epoch": 3.19, "learning_rate": 5.897179435887178e-06, "loss": 0.0219, "step": 10265 }, { "epoch": 3.19, "learning_rate": 5.896779355871175e-06, "loss": 0.0553, "step": 10266 }, { "epoch": 3.19, "learning_rate": 5.8963792758551715e-06, "loss": 0.0502, "step": 10267 }, { "epoch": 3.19, "learning_rate": 5.895979195839169e-06, "loss": 0.0637, "step": 10268 }, { "epoch": 3.19, "learning_rate": 5.895579115823166e-06, "loss": 0.0431, "step": 10269 }, { "epoch": 3.19, "learning_rate": 5.895179035807162e-06, "loss": 0.0486, "step": 10270 }, { "epoch": 3.19, "learning_rate": 5.894778955791158e-06, "loss": 0.0287, "step": 10271 }, { "epoch": 3.19, "learning_rate": 5.894378875775155e-06, "loss": 0.0234, "step": 10272 }, { "epoch": 3.19, "learning_rate": 5.8939787957591525e-06, "loss": 0.1538, "step": 10273 }, { "epoch": 3.19, "learning_rate": 5.893578715743149e-06, "loss": 0.0602, "step": 10274 }, { "epoch": 3.19, "learning_rate": 5.893178635727146e-06, "loss": 0.1283, "step": 10275 }, { "epoch": 3.19, "learning_rate": 5.8927785557111426e-06, "loss": 0.0321, "step": 10276 }, { "epoch": 3.19, "learning_rate": 5.892378475695139e-06, "loss": 0.0627, "step": 10277 }, { "epoch": 3.19, "learning_rate": 5.891978395679137e-06, "loss": 0.0293, "step": 10278 }, { "epoch": 3.19, "learning_rate": 5.8915783156631335e-06, "loss": 0.0689, "step": 10279 }, { "epoch": 3.19, "learning_rate": 5.89117823564713e-06, "loss": 0.0163, "step": 10280 }, { "epoch": 3.19, "learning_rate": 5.890778155631127e-06, "loss": 0.0219, "step": 10281 }, { "epoch": 3.19, "learning_rate": 5.890378075615123e-06, "loss": 0.0421, "step": 10282 }, { "epoch": 3.19, "learning_rate": 5.88997799559912e-06, "loss": 0.0606, "step": 10283 }, { "epoch": 3.19, "learning_rate": 5.889577915583117e-06, "loss": 0.0399, "step": 10284 }, { "epoch": 3.19, "learning_rate": 5.889177835567114e-06, "loss": 0.172, "step": 10285 }, { "epoch": 3.19, "learning_rate": 5.88877775555111e-06, "loss": 0.0398, "step": 10286 }, { "epoch": 3.19, "learning_rate": 5.888377675535108e-06, "loss": 0.046, "step": 10287 }, { "epoch": 3.19, "learning_rate": 5.8879775955191045e-06, "loss": 0.018, "step": 10288 }, { "epoch": 3.19, "learning_rate": 5.887577515503101e-06, "loss": 0.0167, "step": 10289 }, { "epoch": 3.19, "learning_rate": 5.887177435487098e-06, "loss": 0.047, "step": 10290 }, { "epoch": 3.19, "learning_rate": 5.8867773554710946e-06, "loss": 0.0331, "step": 10291 }, { "epoch": 3.19, "learning_rate": 5.886377275455092e-06, "loss": 0.188, "step": 10292 }, { "epoch": 3.19, "learning_rate": 5.885977195439089e-06, "loss": 0.0378, "step": 10293 }, { "epoch": 3.2, "learning_rate": 5.8855771154230855e-06, "loss": 0.0349, "step": 10294 }, { "epoch": 3.2, "learning_rate": 5.885177035407081e-06, "loss": 0.0594, "step": 10295 }, { "epoch": 3.2, "learning_rate": 5.884776955391078e-06, "loss": 0.0418, "step": 10296 }, { "epoch": 3.2, "learning_rate": 5.8843768753750755e-06, "loss": 0.0534, "step": 10297 }, { "epoch": 3.2, "learning_rate": 5.883976795359072e-06, "loss": 0.0105, "step": 10298 }, { "epoch": 3.2, "learning_rate": 5.883576715343069e-06, "loss": 0.0169, "step": 10299 }, { "epoch": 3.2, "learning_rate": 5.883176635327066e-06, "loss": 0.0406, "step": 10300 }, { "epoch": 3.2, "learning_rate": 5.882776555311063e-06, "loss": 0.0382, "step": 10301 }, { "epoch": 3.2, "learning_rate": 5.88237647529506e-06, "loss": 0.037, "step": 10302 }, { "epoch": 3.2, "learning_rate": 5.8819763952790565e-06, "loss": 0.0614, "step": 10303 }, { "epoch": 3.2, "learning_rate": 5.881576315263053e-06, "loss": 0.0143, "step": 10304 }, { "epoch": 3.2, "learning_rate": 5.88117623524705e-06, "loss": 0.043, "step": 10305 }, { "epoch": 3.2, "learning_rate": 5.880776155231047e-06, "loss": 0.0642, "step": 10306 }, { "epoch": 3.2, "learning_rate": 5.880376075215043e-06, "loss": 0.0268, "step": 10307 }, { "epoch": 3.2, "learning_rate": 5.87997599519904e-06, "loss": 0.0251, "step": 10308 }, { "epoch": 3.2, "learning_rate": 5.879575915183037e-06, "loss": 0.0271, "step": 10309 }, { "epoch": 3.2, "learning_rate": 5.879175835167033e-06, "loss": 0.0235, "step": 10310 }, { "epoch": 3.2, "learning_rate": 5.878775755151031e-06, "loss": 0.0632, "step": 10311 }, { "epoch": 3.2, "learning_rate": 5.8783756751350275e-06, "loss": 0.0396, "step": 10312 }, { "epoch": 3.2, "learning_rate": 5.877975595119024e-06, "loss": 0.0207, "step": 10313 }, { "epoch": 3.2, "learning_rate": 5.877575515103021e-06, "loss": 0.0197, "step": 10314 }, { "epoch": 3.2, "learning_rate": 5.877175435087018e-06, "loss": 0.1301, "step": 10315 }, { "epoch": 3.2, "learning_rate": 5.876775355071015e-06, "loss": 0.0179, "step": 10316 }, { "epoch": 3.2, "learning_rate": 5.876375275055012e-06, "loss": 0.0791, "step": 10317 }, { "epoch": 3.2, "learning_rate": 5.8759751950390085e-06, "loss": 0.0191, "step": 10318 }, { "epoch": 3.2, "learning_rate": 5.875575115023006e-06, "loss": 0.0358, "step": 10319 }, { "epoch": 3.2, "learning_rate": 5.875175035007001e-06, "loss": 0.0079, "step": 10320 }, { "epoch": 3.2, "learning_rate": 5.8747749549909985e-06, "loss": 0.0176, "step": 10321 }, { "epoch": 3.2, "learning_rate": 5.874374874974995e-06, "loss": 0.0631, "step": 10322 }, { "epoch": 3.2, "learning_rate": 5.873974794958992e-06, "loss": 0.0267, "step": 10323 }, { "epoch": 3.2, "learning_rate": 5.873574714942989e-06, "loss": 0.2217, "step": 10324 }, { "epoch": 3.2, "learning_rate": 5.873174634926986e-06, "loss": 0.0228, "step": 10325 }, { "epoch": 3.2, "learning_rate": 5.872774554910983e-06, "loss": 0.0375, "step": 10326 }, { "epoch": 3.21, "learning_rate": 5.8723744748949795e-06, "loss": 0.0209, "step": 10327 }, { "epoch": 3.21, "learning_rate": 5.871974394878976e-06, "loss": 0.036, "step": 10328 }, { "epoch": 3.21, "learning_rate": 5.871574314862974e-06, "loss": 0.0402, "step": 10329 }, { "epoch": 3.21, "learning_rate": 5.87117423484697e-06, "loss": 0.0332, "step": 10330 }, { "epoch": 3.21, "learning_rate": 5.870774154830967e-06, "loss": 0.0598, "step": 10331 }, { "epoch": 3.21, "learning_rate": 5.870374074814964e-06, "loss": 0.0272, "step": 10332 }, { "epoch": 3.21, "learning_rate": 5.86997399479896e-06, "loss": 0.0363, "step": 10333 }, { "epoch": 3.21, "learning_rate": 5.869573914782956e-06, "loss": 0.0394, "step": 10334 }, { "epoch": 3.21, "learning_rate": 5.869173834766954e-06, "loss": 0.0443, "step": 10335 }, { "epoch": 3.21, "learning_rate": 5.8687737547509505e-06, "loss": 0.0246, "step": 10336 }, { "epoch": 3.21, "learning_rate": 5.868373674734947e-06, "loss": 0.043, "step": 10337 }, { "epoch": 3.21, "learning_rate": 5.867973594718944e-06, "loss": 0.0411, "step": 10338 }, { "epoch": 3.21, "learning_rate": 5.8675735147029414e-06, "loss": 0.0841, "step": 10339 }, { "epoch": 3.21, "learning_rate": 5.867173434686938e-06, "loss": 0.0241, "step": 10340 }, { "epoch": 3.21, "learning_rate": 5.866773354670935e-06, "loss": 0.017, "step": 10341 }, { "epoch": 3.21, "learning_rate": 5.8663732746549315e-06, "loss": 0.0416, "step": 10342 }, { "epoch": 3.21, "learning_rate": 5.865973194638929e-06, "loss": 0.0479, "step": 10343 }, { "epoch": 3.21, "learning_rate": 5.865573114622926e-06, "loss": 0.0275, "step": 10344 }, { "epoch": 3.21, "learning_rate": 5.8651730346069215e-06, "loss": 0.0371, "step": 10345 }, { "epoch": 3.21, "learning_rate": 5.864772954590918e-06, "loss": 0.0512, "step": 10346 }, { "epoch": 3.21, "learning_rate": 5.864372874574915e-06, "loss": 0.0775, "step": 10347 }, { "epoch": 3.21, "learning_rate": 5.863972794558912e-06, "loss": 0.0546, "step": 10348 }, { "epoch": 3.21, "learning_rate": 5.863572714542909e-06, "loss": 0.0839, "step": 10349 }, { "epoch": 3.21, "learning_rate": 5.863172634526906e-06, "loss": 0.109, "step": 10350 }, { "epoch": 3.21, "learning_rate": 5.8627725545109025e-06, "loss": 0.0395, "step": 10351 }, { "epoch": 3.21, "learning_rate": 5.862372474494899e-06, "loss": 0.041, "step": 10352 }, { "epoch": 3.21, "learning_rate": 5.861972394478897e-06, "loss": 0.0581, "step": 10353 }, { "epoch": 3.21, "learning_rate": 5.861572314462893e-06, "loss": 0.1016, "step": 10354 }, { "epoch": 3.21, "learning_rate": 5.86117223444689e-06, "loss": 0.048, "step": 10355 }, { "epoch": 3.21, "learning_rate": 5.860772154430887e-06, "loss": 0.0383, "step": 10356 }, { "epoch": 3.21, "learning_rate": 5.860372074414884e-06, "loss": 0.0488, "step": 10357 }, { "epoch": 3.21, "learning_rate": 5.85997199439888e-06, "loss": 0.0188, "step": 10358 }, { "epoch": 3.22, "learning_rate": 5.859571914382877e-06, "loss": 0.1535, "step": 10359 }, { "epoch": 3.22, "learning_rate": 5.8591718343668735e-06, "loss": 0.0649, "step": 10360 }, { "epoch": 3.22, "learning_rate": 5.85877175435087e-06, "loss": 0.0272, "step": 10361 }, { "epoch": 3.22, "learning_rate": 5.858371674334867e-06, "loss": 0.032, "step": 10362 }, { "epoch": 3.22, "learning_rate": 5.8579715943188644e-06, "loss": 0.0373, "step": 10363 }, { "epoch": 3.22, "learning_rate": 5.857571514302861e-06, "loss": 0.033, "step": 10364 }, { "epoch": 3.22, "learning_rate": 5.857171434286858e-06, "loss": 0.0252, "step": 10365 }, { "epoch": 3.22, "learning_rate": 5.8567713542708545e-06, "loss": 0.0559, "step": 10366 }, { "epoch": 3.22, "learning_rate": 5.856371274254852e-06, "loss": 0.0373, "step": 10367 }, { "epoch": 3.22, "learning_rate": 5.855971194238849e-06, "loss": 0.1461, "step": 10368 }, { "epoch": 3.22, "learning_rate": 5.855571114222845e-06, "loss": 0.0345, "step": 10369 }, { "epoch": 3.22, "learning_rate": 5.855171034206841e-06, "loss": 0.0426, "step": 10370 }, { "epoch": 3.22, "learning_rate": 5.854770954190838e-06, "loss": 0.0369, "step": 10371 }, { "epoch": 3.22, "learning_rate": 5.8543708741748355e-06, "loss": 0.0663, "step": 10372 }, { "epoch": 3.22, "learning_rate": 5.853970794158832e-06, "loss": 0.0426, "step": 10373 }, { "epoch": 3.22, "learning_rate": 5.853570714142829e-06, "loss": 0.0187, "step": 10374 }, { "epoch": 3.22, "learning_rate": 5.8531706341268255e-06, "loss": 0.0237, "step": 10375 }, { "epoch": 3.22, "learning_rate": 5.852770554110822e-06, "loss": 0.0385, "step": 10376 }, { "epoch": 3.22, "learning_rate": 5.85237047409482e-06, "loss": 0.056, "step": 10377 }, { "epoch": 3.22, "learning_rate": 5.8519703940788164e-06, "loss": 0.0424, "step": 10378 }, { "epoch": 3.22, "learning_rate": 5.851570314062813e-06, "loss": 0.0235, "step": 10379 }, { "epoch": 3.22, "learning_rate": 5.85117023404681e-06, "loss": 0.0239, "step": 10380 }, { "epoch": 3.22, "learning_rate": 5.850770154030807e-06, "loss": 0.0637, "step": 10381 }, { "epoch": 3.22, "learning_rate": 5.850370074014804e-06, "loss": 0.0316, "step": 10382 }, { "epoch": 3.22, "learning_rate": 5.8499699939988e-06, "loss": 0.0293, "step": 10383 }, { "epoch": 3.22, "learning_rate": 5.8495699139827965e-06, "loss": 0.032, "step": 10384 }, { "epoch": 3.22, "learning_rate": 5.849169833966793e-06, "loss": 0.0494, "step": 10385 }, { "epoch": 3.22, "learning_rate": 5.848769753950791e-06, "loss": 0.0848, "step": 10386 }, { "epoch": 3.22, "learning_rate": 5.8483696739347874e-06, "loss": 0.016, "step": 10387 }, { "epoch": 3.22, "learning_rate": 5.847969593918784e-06, "loss": 0.075, "step": 10388 }, { "epoch": 3.22, "learning_rate": 5.847569513902781e-06, "loss": 0.065, "step": 10389 }, { "epoch": 3.22, "learning_rate": 5.847169433886778e-06, "loss": 0.0256, "step": 10390 }, { "epoch": 3.23, "learning_rate": 5.846769353870775e-06, "loss": 0.0369, "step": 10391 }, { "epoch": 3.23, "learning_rate": 5.846369273854772e-06, "loss": 0.0194, "step": 10392 }, { "epoch": 3.23, "learning_rate": 5.845969193838768e-06, "loss": 0.0383, "step": 10393 }, { "epoch": 3.23, "learning_rate": 5.845569113822765e-06, "loss": 0.0451, "step": 10394 }, { "epoch": 3.23, "learning_rate": 5.845169033806763e-06, "loss": 0.1269, "step": 10395 }, { "epoch": 3.23, "learning_rate": 5.8447689537907585e-06, "loss": 0.1192, "step": 10396 }, { "epoch": 3.23, "learning_rate": 5.844368873774755e-06, "loss": 0.0272, "step": 10397 }, { "epoch": 3.23, "learning_rate": 5.843968793758752e-06, "loss": 0.0693, "step": 10398 }, { "epoch": 3.23, "learning_rate": 5.8435687137427485e-06, "loss": 0.0757, "step": 10399 }, { "epoch": 3.23, "learning_rate": 5.843168633726746e-06, "loss": 0.1199, "step": 10400 }, { "epoch": 3.23, "learning_rate": 5.842768553710743e-06, "loss": 0.0584, "step": 10401 }, { "epoch": 3.23, "learning_rate": 5.8423684736947394e-06, "loss": 0.1714, "step": 10402 }, { "epoch": 3.23, "learning_rate": 5.841968393678736e-06, "loss": 0.1221, "step": 10403 }, { "epoch": 3.23, "learning_rate": 5.841568313662734e-06, "loss": 0.0317, "step": 10404 }, { "epoch": 3.23, "learning_rate": 5.84116823364673e-06, "loss": 0.0234, "step": 10405 }, { "epoch": 3.23, "learning_rate": 5.840768153630727e-06, "loss": 0.0457, "step": 10406 }, { "epoch": 3.23, "learning_rate": 5.840368073614724e-06, "loss": 0.0262, "step": 10407 }, { "epoch": 3.23, "learning_rate": 5.8399679935987196e-06, "loss": 0.1288, "step": 10408 }, { "epoch": 3.23, "learning_rate": 5.839567913582716e-06, "loss": 0.0358, "step": 10409 }, { "epoch": 3.23, "learning_rate": 5.839167833566714e-06, "loss": 0.0453, "step": 10410 }, { "epoch": 3.23, "learning_rate": 5.8387677535507105e-06, "loss": 0.0354, "step": 10411 }, { "epoch": 3.23, "learning_rate": 5.838367673534707e-06, "loss": 0.0388, "step": 10412 }, { "epoch": 3.23, "learning_rate": 5.837967593518704e-06, "loss": 0.0382, "step": 10413 }, { "epoch": 3.23, "learning_rate": 5.837567513502701e-06, "loss": 0.0351, "step": 10414 }, { "epoch": 3.23, "learning_rate": 5.837167433486698e-06, "loss": 0.0427, "step": 10415 }, { "epoch": 3.23, "learning_rate": 5.836767353470695e-06, "loss": 0.0208, "step": 10416 }, { "epoch": 3.23, "learning_rate": 5.8363672734546914e-06, "loss": 0.0282, "step": 10417 }, { "epoch": 3.23, "learning_rate": 5.835967193438689e-06, "loss": 0.0877, "step": 10418 }, { "epoch": 3.23, "learning_rate": 5.835567113422686e-06, "loss": 0.0377, "step": 10419 }, { "epoch": 3.23, "learning_rate": 5.835167033406682e-06, "loss": 0.0936, "step": 10420 }, { "epoch": 3.23, "learning_rate": 5.834766953390678e-06, "loss": 0.0392, "step": 10421 }, { "epoch": 3.23, "learning_rate": 5.834366873374675e-06, "loss": 0.0385, "step": 10422 }, { "epoch": 3.24, "learning_rate": 5.8339667933586715e-06, "loss": 0.0178, "step": 10423 }, { "epoch": 3.24, "learning_rate": 5.833566713342669e-06, "loss": 0.0809, "step": 10424 }, { "epoch": 3.24, "learning_rate": 5.833166633326666e-06, "loss": 0.0358, "step": 10425 }, { "epoch": 3.24, "learning_rate": 5.8327665533106624e-06, "loss": 0.0284, "step": 10426 }, { "epoch": 3.24, "learning_rate": 5.832366473294659e-06, "loss": 0.0662, "step": 10427 }, { "epoch": 3.24, "learning_rate": 5.831966393278657e-06, "loss": 0.0588, "step": 10428 }, { "epoch": 3.24, "learning_rate": 5.831566313262653e-06, "loss": 0.0362, "step": 10429 }, { "epoch": 3.24, "learning_rate": 5.83116623324665e-06, "loss": 0.0789, "step": 10430 }, { "epoch": 3.24, "learning_rate": 5.830766153230647e-06, "loss": 0.1127, "step": 10431 }, { "epoch": 3.24, "learning_rate": 5.830366073214644e-06, "loss": 0.0456, "step": 10432 }, { "epoch": 3.24, "learning_rate": 5.829965993198639e-06, "loss": 0.1526, "step": 10433 }, { "epoch": 3.24, "learning_rate": 5.829565913182637e-06, "loss": 0.0651, "step": 10434 }, { "epoch": 3.24, "learning_rate": 5.8291658331666335e-06, "loss": 0.024, "step": 10435 }, { "epoch": 3.24, "learning_rate": 5.82876575315063e-06, "loss": 0.1049, "step": 10436 }, { "epoch": 3.24, "learning_rate": 5.828365673134627e-06, "loss": 0.1222, "step": 10437 }, { "epoch": 3.24, "learning_rate": 5.827965593118624e-06, "loss": 0.1724, "step": 10438 }, { "epoch": 3.24, "learning_rate": 5.827565513102621e-06, "loss": 0.0521, "step": 10439 }, { "epoch": 3.24, "learning_rate": 5.827165433086618e-06, "loss": 0.0333, "step": 10440 }, { "epoch": 3.24, "learning_rate": 5.8267653530706144e-06, "loss": 0.0326, "step": 10441 }, { "epoch": 3.24, "learning_rate": 5.826365273054612e-06, "loss": 0.0133, "step": 10442 }, { "epoch": 3.24, "learning_rate": 5.825965193038609e-06, "loss": 0.0539, "step": 10443 }, { "epoch": 3.24, "learning_rate": 5.825565113022605e-06, "loss": 0.1134, "step": 10444 }, { "epoch": 3.24, "learning_rate": 5.825165033006602e-06, "loss": 0.0121, "step": 10445 }, { "epoch": 3.24, "learning_rate": 5.824764952990598e-06, "loss": 0.0331, "step": 10446 }, { "epoch": 3.24, "learning_rate": 5.8243648729745946e-06, "loss": 0.0334, "step": 10447 }, { "epoch": 3.24, "learning_rate": 5.823964792958592e-06, "loss": 0.1051, "step": 10448 }, { "epoch": 3.24, "learning_rate": 5.823564712942589e-06, "loss": 0.0528, "step": 10449 }, { "epoch": 3.24, "learning_rate": 5.8231646329265855e-06, "loss": 0.0564, "step": 10450 }, { "epoch": 3.24, "learning_rate": 5.822764552910582e-06, "loss": 0.0628, "step": 10451 }, { "epoch": 3.24, "learning_rate": 5.82236447289458e-06, "loss": 0.0613, "step": 10452 }, { "epoch": 3.24, "learning_rate": 5.821964392878576e-06, "loss": 0.0117, "step": 10453 }, { "epoch": 3.24, "learning_rate": 5.821564312862573e-06, "loss": 0.0127, "step": 10454 }, { "epoch": 3.25, "learning_rate": 5.82116423284657e-06, "loss": 0.0174, "step": 10455 }, { "epoch": 3.25, "learning_rate": 5.820764152830567e-06, "loss": 0.0172, "step": 10456 }, { "epoch": 3.25, "learning_rate": 5.820364072814564e-06, "loss": 0.0107, "step": 10457 }, { "epoch": 3.25, "learning_rate": 5.819963992798561e-06, "loss": 0.0465, "step": 10458 }, { "epoch": 3.25, "learning_rate": 5.8195639127825565e-06, "loss": 0.0369, "step": 10459 }, { "epoch": 3.25, "learning_rate": 5.819163832766553e-06, "loss": 0.1124, "step": 10460 }, { "epoch": 3.25, "learning_rate": 5.81876375275055e-06, "loss": 0.0434, "step": 10461 }, { "epoch": 3.25, "learning_rate": 5.818363672734547e-06, "loss": 0.0563, "step": 10462 }, { "epoch": 3.25, "learning_rate": 5.817963592718544e-06, "loss": 0.0391, "step": 10463 }, { "epoch": 3.25, "learning_rate": 5.817563512702541e-06, "loss": 0.0333, "step": 10464 }, { "epoch": 3.25, "learning_rate": 5.8171634326865374e-06, "loss": 0.0605, "step": 10465 }, { "epoch": 3.25, "learning_rate": 5.816763352670535e-06, "loss": 0.0712, "step": 10466 }, { "epoch": 3.25, "learning_rate": 5.816363272654532e-06, "loss": 0.0283, "step": 10467 }, { "epoch": 3.25, "learning_rate": 5.815963192638528e-06, "loss": 0.0504, "step": 10468 }, { "epoch": 3.25, "learning_rate": 5.815563112622525e-06, "loss": 0.0295, "step": 10469 }, { "epoch": 3.25, "learning_rate": 5.8151630326065226e-06, "loss": 0.0469, "step": 10470 }, { "epoch": 3.25, "learning_rate": 5.814762952590518e-06, "loss": 0.1096, "step": 10471 }, { "epoch": 3.25, "learning_rate": 5.814362872574515e-06, "loss": 0.0917, "step": 10472 }, { "epoch": 3.25, "learning_rate": 5.813962792558512e-06, "loss": 0.0299, "step": 10473 }, { "epoch": 3.25, "learning_rate": 5.8135627125425085e-06, "loss": 0.0388, "step": 10474 }, { "epoch": 3.25, "learning_rate": 5.813162632526506e-06, "loss": 0.0652, "step": 10475 }, { "epoch": 3.25, "learning_rate": 5.812762552510503e-06, "loss": 0.0448, "step": 10476 }, { "epoch": 3.25, "learning_rate": 5.812362472494499e-06, "loss": 0.0394, "step": 10477 }, { "epoch": 3.25, "learning_rate": 5.811962392478496e-06, "loss": 0.0733, "step": 10478 }, { "epoch": 3.25, "learning_rate": 5.811562312462493e-06, "loss": 0.0263, "step": 10479 }, { "epoch": 3.25, "learning_rate": 5.81116223244649e-06, "loss": 0.043, "step": 10480 }, { "epoch": 3.25, "learning_rate": 5.810762152430487e-06, "loss": 0.0355, "step": 10481 }, { "epoch": 3.25, "learning_rate": 5.810362072414484e-06, "loss": 0.0364, "step": 10482 }, { "epoch": 3.25, "learning_rate": 5.80996199239848e-06, "loss": 0.0456, "step": 10483 }, { "epoch": 3.25, "learning_rate": 5.809561912382476e-06, "loss": 0.0518, "step": 10484 }, { "epoch": 3.25, "learning_rate": 5.809161832366474e-06, "loss": 0.0434, "step": 10485 }, { "epoch": 3.25, "learning_rate": 5.80876175235047e-06, "loss": 0.0173, "step": 10486 }, { "epoch": 3.25, "learning_rate": 5.808361672334467e-06, "loss": 0.1091, "step": 10487 }, { "epoch": 3.26, "learning_rate": 5.807961592318464e-06, "loss": 0.044, "step": 10488 }, { "epoch": 3.26, "learning_rate": 5.807561512302461e-06, "loss": 0.0373, "step": 10489 }, { "epoch": 3.26, "learning_rate": 5.807161432286458e-06, "loss": 0.0164, "step": 10490 }, { "epoch": 3.26, "learning_rate": 5.806761352270455e-06, "loss": 0.0286, "step": 10491 }, { "epoch": 3.26, "learning_rate": 5.806361272254451e-06, "loss": 0.0451, "step": 10492 }, { "epoch": 3.26, "learning_rate": 5.805961192238448e-06, "loss": 0.0288, "step": 10493 }, { "epoch": 3.26, "learning_rate": 5.805561112222446e-06, "loss": 0.0891, "step": 10494 }, { "epoch": 3.26, "learning_rate": 5.805161032206442e-06, "loss": 0.0445, "step": 10495 }, { "epoch": 3.26, "learning_rate": 5.804760952190439e-06, "loss": 0.1348, "step": 10496 }, { "epoch": 3.26, "learning_rate": 5.804360872174435e-06, "loss": 0.0549, "step": 10497 }, { "epoch": 3.26, "learning_rate": 5.8039607921584315e-06, "loss": 0.042, "step": 10498 }, { "epoch": 3.26, "learning_rate": 5.803560712142429e-06, "loss": 0.0294, "step": 10499 }, { "epoch": 3.26, "learning_rate": 5.803160632126426e-06, "loss": 0.1787, "step": 10500 }, { "epoch": 3.26, "learning_rate": 5.802760552110422e-06, "loss": 0.1366, "step": 10501 }, { "epoch": 3.26, "learning_rate": 5.802360472094419e-06, "loss": 0.0758, "step": 10502 }, { "epoch": 3.26, "learning_rate": 5.801960392078417e-06, "loss": 0.0346, "step": 10503 }, { "epoch": 3.26, "learning_rate": 5.801560312062413e-06, "loss": 0.0307, "step": 10504 }, { "epoch": 3.26, "learning_rate": 5.801560312062413e-06, "loss": 0.0366, "step": 10505 }, { "epoch": 3.26, "learning_rate": 5.80116023204641e-06, "loss": 0.0847, "step": 10506 }, { "epoch": 3.26, "learning_rate": 5.800760152030407e-06, "loss": 0.0338, "step": 10507 }, { "epoch": 3.26, "learning_rate": 5.800360072014403e-06, "loss": 0.0464, "step": 10508 }, { "epoch": 3.26, "learning_rate": 5.799959991998401e-06, "loss": 0.0379, "step": 10509 }, { "epoch": 3.26, "learning_rate": 5.799559911982397e-06, "loss": 0.0367, "step": 10510 }, { "epoch": 3.26, "learning_rate": 5.799159831966393e-06, "loss": 0.028, "step": 10511 }, { "epoch": 3.26, "learning_rate": 5.79875975195039e-06, "loss": 0.0442, "step": 10512 }, { "epoch": 3.26, "learning_rate": 5.798359671934387e-06, "loss": 0.0304, "step": 10513 }, { "epoch": 3.26, "learning_rate": 5.797959591918384e-06, "loss": 0.0307, "step": 10514 }, { "epoch": 3.26, "learning_rate": 5.797559511902381e-06, "loss": 0.0657, "step": 10515 }, { "epoch": 3.26, "learning_rate": 5.797159431886378e-06, "loss": 0.0441, "step": 10516 }, { "epoch": 3.26, "learning_rate": 5.796759351870374e-06, "loss": 0.0278, "step": 10517 }, { "epoch": 3.26, "learning_rate": 5.796359271854372e-06, "loss": 0.0403, "step": 10518 }, { "epoch": 3.26, "learning_rate": 5.795959191838369e-06, "loss": 0.0856, "step": 10519 }, { "epoch": 3.27, "learning_rate": 5.795559111822365e-06, "loss": 0.14, "step": 10520 }, { "epoch": 3.27, "learning_rate": 5.795159031806362e-06, "loss": 0.0232, "step": 10521 }, { "epoch": 3.27, "learning_rate": 5.794758951790359e-06, "loss": 0.0253, "step": 10522 }, { "epoch": 3.27, "learning_rate": 5.7943588717743545e-06, "loss": 0.0446, "step": 10523 }, { "epoch": 3.27, "learning_rate": 5.793958791758352e-06, "loss": 0.0519, "step": 10524 }, { "epoch": 3.27, "learning_rate": 5.793558711742349e-06, "loss": 0.0731, "step": 10525 }, { "epoch": 3.27, "learning_rate": 5.793158631726345e-06, "loss": 0.0459, "step": 10526 }, { "epoch": 3.27, "learning_rate": 5.792758551710342e-06, "loss": 0.0321, "step": 10527 }, { "epoch": 3.27, "learning_rate": 5.79235847169434e-06, "loss": 0.0353, "step": 10528 }, { "epoch": 3.27, "learning_rate": 5.791958391678336e-06, "loss": 0.0252, "step": 10529 }, { "epoch": 3.27, "learning_rate": 5.791558311662333e-06, "loss": 0.0524, "step": 10530 }, { "epoch": 3.27, "learning_rate": 5.79115823164633e-06, "loss": 0.1414, "step": 10531 }, { "epoch": 3.27, "learning_rate": 5.790758151630327e-06, "loss": 0.0247, "step": 10532 }, { "epoch": 3.27, "learning_rate": 5.790358071614324e-06, "loss": 0.0195, "step": 10533 }, { "epoch": 3.27, "learning_rate": 5.789957991598321e-06, "loss": 0.1041, "step": 10534 }, { "epoch": 3.27, "learning_rate": 5.7895579115823164e-06, "loss": 0.0503, "step": 10535 }, { "epoch": 3.27, "learning_rate": 5.789157831566313e-06, "loss": 0.0513, "step": 10536 }, { "epoch": 3.27, "learning_rate": 5.78875775155031e-06, "loss": 0.0348, "step": 10537 }, { "epoch": 3.27, "learning_rate": 5.788357671534307e-06, "loss": 0.0666, "step": 10538 }, { "epoch": 3.27, "learning_rate": 5.787957591518304e-06, "loss": 0.0579, "step": 10539 }, { "epoch": 3.27, "learning_rate": 5.787557511502301e-06, "loss": 0.0375, "step": 10540 }, { "epoch": 3.27, "learning_rate": 5.787157431486297e-06, "loss": 0.0621, "step": 10541 }, { "epoch": 3.27, "learning_rate": 5.786757351470295e-06, "loss": 0.0472, "step": 10542 }, { "epoch": 3.27, "learning_rate": 5.786357271454292e-06, "loss": 0.1035, "step": 10543 }, { "epoch": 3.27, "learning_rate": 5.785957191438288e-06, "loss": 0.041, "step": 10544 }, { "epoch": 3.27, "learning_rate": 5.785557111422285e-06, "loss": 0.0561, "step": 10545 }, { "epoch": 3.27, "learning_rate": 5.7851570314062825e-06, "loss": 0.0375, "step": 10546 }, { "epoch": 3.27, "learning_rate": 5.784756951390279e-06, "loss": 0.0731, "step": 10547 }, { "epoch": 3.27, "learning_rate": 5.784356871374275e-06, "loss": 0.0593, "step": 10548 }, { "epoch": 3.27, "learning_rate": 5.783956791358272e-06, "loss": 0.0262, "step": 10549 }, { "epoch": 3.27, "learning_rate": 5.783556711342268e-06, "loss": 0.066, "step": 10550 }, { "epoch": 3.27, "learning_rate": 5.783156631326265e-06, "loss": 0.0536, "step": 10551 }, { "epoch": 3.28, "learning_rate": 5.782756551310263e-06, "loss": 0.0721, "step": 10552 }, { "epoch": 3.28, "learning_rate": 5.782356471294259e-06, "loss": 0.0264, "step": 10553 }, { "epoch": 3.28, "learning_rate": 5.781956391278256e-06, "loss": 0.0313, "step": 10554 }, { "epoch": 3.28, "learning_rate": 5.781556311262253e-06, "loss": 0.0413, "step": 10555 }, { "epoch": 3.28, "learning_rate": 5.78115623124625e-06, "loss": 0.0756, "step": 10556 }, { "epoch": 3.28, "learning_rate": 5.780756151230247e-06, "loss": 0.046, "step": 10557 }, { "epoch": 3.28, "learning_rate": 5.780356071214244e-06, "loss": 0.0716, "step": 10558 }, { "epoch": 3.28, "learning_rate": 5.77995599119824e-06, "loss": 0.0309, "step": 10559 }, { "epoch": 3.28, "learning_rate": 5.779555911182238e-06, "loss": 0.0415, "step": 10560 }, { "epoch": 3.28, "learning_rate": 5.779155831166234e-06, "loss": 0.0232, "step": 10561 }, { "epoch": 3.28, "learning_rate": 5.77875575115023e-06, "loss": 0.0464, "step": 10562 }, { "epoch": 3.28, "learning_rate": 5.778355671134227e-06, "loss": 0.0558, "step": 10563 }, { "epoch": 3.28, "learning_rate": 5.777955591118224e-06, "loss": 0.071, "step": 10564 }, { "epoch": 3.28, "learning_rate": 5.77755551110222e-06, "loss": 0.05, "step": 10565 }, { "epoch": 3.28, "learning_rate": 5.777155431086218e-06, "loss": 0.072, "step": 10566 }, { "epoch": 3.28, "learning_rate": 5.776755351070215e-06, "loss": 0.0392, "step": 10567 }, { "epoch": 3.28, "learning_rate": 5.776355271054211e-06, "loss": 0.0498, "step": 10568 }, { "epoch": 3.28, "learning_rate": 5.775955191038208e-06, "loss": 0.0489, "step": 10569 }, { "epoch": 3.28, "learning_rate": 5.7755551110222055e-06, "loss": 0.0118, "step": 10570 }, { "epoch": 3.28, "learning_rate": 5.775155031006202e-06, "loss": 0.0116, "step": 10571 }, { "epoch": 3.28, "learning_rate": 5.774754950990199e-06, "loss": 0.0433, "step": 10572 }, { "epoch": 3.28, "learning_rate": 5.774354870974195e-06, "loss": 0.0075, "step": 10573 }, { "epoch": 3.28, "learning_rate": 5.7739547909581914e-06, "loss": 0.0982, "step": 10574 }, { "epoch": 3.28, "learning_rate": 5.773554710942189e-06, "loss": 0.064, "step": 10575 }, { "epoch": 3.28, "learning_rate": 5.773154630926186e-06, "loss": 0.0403, "step": 10576 }, { "epoch": 3.28, "learning_rate": 5.772754550910182e-06, "loss": 0.014, "step": 10577 }, { "epoch": 3.28, "learning_rate": 5.772354470894179e-06, "loss": 0.0196, "step": 10578 }, { "epoch": 3.28, "learning_rate": 5.771954390878176e-06, "loss": 0.0297, "step": 10579 }, { "epoch": 3.28, "learning_rate": 5.771554310862173e-06, "loss": 0.0225, "step": 10580 }, { "epoch": 3.28, "learning_rate": 5.77115423084617e-06, "loss": 0.0328, "step": 10581 }, { "epoch": 3.28, "learning_rate": 5.770754150830167e-06, "loss": 0.0271, "step": 10582 }, { "epoch": 3.28, "learning_rate": 5.770354070814163e-06, "loss": 0.0177, "step": 10583 }, { "epoch": 3.29, "learning_rate": 5.769953990798161e-06, "loss": 0.0218, "step": 10584 }, { "epoch": 3.29, "learning_rate": 5.7695539107821575e-06, "loss": 0.0235, "step": 10585 }, { "epoch": 3.29, "learning_rate": 5.769153830766153e-06, "loss": 0.0417, "step": 10586 }, { "epoch": 3.29, "learning_rate": 5.76875375075015e-06, "loss": 0.0317, "step": 10587 }, { "epoch": 3.29, "learning_rate": 5.768353670734147e-06, "loss": 0.0427, "step": 10588 }, { "epoch": 3.29, "learning_rate": 5.767953590718144e-06, "loss": 0.1654, "step": 10589 }, { "epoch": 3.29, "learning_rate": 5.767553510702141e-06, "loss": 0.0685, "step": 10590 }, { "epoch": 3.29, "learning_rate": 5.767153430686138e-06, "loss": 0.0348, "step": 10591 }, { "epoch": 3.29, "learning_rate": 5.766753350670134e-06, "loss": 0.0366, "step": 10592 }, { "epoch": 3.29, "learning_rate": 5.766353270654131e-06, "loss": 0.0535, "step": 10593 }, { "epoch": 3.29, "learning_rate": 5.7659531906381285e-06, "loss": 0.0304, "step": 10594 }, { "epoch": 3.29, "learning_rate": 5.765553110622125e-06, "loss": 0.0399, "step": 10595 }, { "epoch": 3.29, "learning_rate": 5.765153030606122e-06, "loss": 0.0766, "step": 10596 }, { "epoch": 3.29, "learning_rate": 5.764752950590119e-06, "loss": 0.0456, "step": 10597 }, { "epoch": 3.29, "learning_rate": 5.7643528705741144e-06, "loss": 0.0507, "step": 10598 }, { "epoch": 3.29, "learning_rate": 5.763952790558112e-06, "loss": 0.0402, "step": 10599 }, { "epoch": 3.29, "learning_rate": 5.763552710542109e-06, "loss": 0.025, "step": 10600 }, { "epoch": 3.29, "learning_rate": 5.763152630526105e-06, "loss": 0.0393, "step": 10601 }, { "epoch": 3.29, "learning_rate": 5.762752550510102e-06, "loss": 0.0219, "step": 10602 }, { "epoch": 3.29, "learning_rate": 5.7623524704940996e-06, "loss": 0.0385, "step": 10603 }, { "epoch": 3.29, "learning_rate": 5.761952390478096e-06, "loss": 0.041, "step": 10604 }, { "epoch": 3.29, "learning_rate": 5.761552310462093e-06, "loss": 0.033, "step": 10605 }, { "epoch": 3.29, "learning_rate": 5.76115223044609e-06, "loss": 0.0452, "step": 10606 }, { "epoch": 3.29, "learning_rate": 5.760752150430087e-06, "loss": 0.0994, "step": 10607 }, { "epoch": 3.29, "learning_rate": 5.760352070414084e-06, "loss": 0.0494, "step": 10608 }, { "epoch": 3.29, "learning_rate": 5.7599519903980805e-06, "loss": 0.0689, "step": 10609 }, { "epoch": 3.29, "learning_rate": 5.759551910382077e-06, "loss": 0.0337, "step": 10610 }, { "epoch": 3.29, "learning_rate": 5.759151830366073e-06, "loss": 0.0836, "step": 10611 }, { "epoch": 3.29, "learning_rate": 5.75875175035007e-06, "loss": 0.0639, "step": 10612 }, { "epoch": 3.29, "learning_rate": 5.758351670334067e-06, "loss": 0.0358, "step": 10613 }, { "epoch": 3.29, "learning_rate": 5.757951590318064e-06, "loss": 0.0424, "step": 10614 }, { "epoch": 3.29, "learning_rate": 5.757551510302061e-06, "loss": 0.0363, "step": 10615 }, { "epoch": 3.29, "learning_rate": 5.757151430286057e-06, "loss": 0.0285, "step": 10616 }, { "epoch": 3.3, "learning_rate": 5.756751350270055e-06, "loss": 0.0582, "step": 10617 }, { "epoch": 3.3, "learning_rate": 5.7563512702540516e-06, "loss": 0.0557, "step": 10618 }, { "epoch": 3.3, "learning_rate": 5.755951190238048e-06, "loss": 0.0364, "step": 10619 }, { "epoch": 3.3, "learning_rate": 5.755551110222045e-06, "loss": 0.0908, "step": 10620 }, { "epoch": 3.3, "learning_rate": 5.7551510302060425e-06, "loss": 0.0281, "step": 10621 }, { "epoch": 3.3, "learning_rate": 5.754750950190039e-06, "loss": 0.0183, "step": 10622 }, { "epoch": 3.3, "learning_rate": 5.754350870174036e-06, "loss": 0.0186, "step": 10623 }, { "epoch": 3.3, "learning_rate": 5.753950790158032e-06, "loss": 0.0239, "step": 10624 }, { "epoch": 3.3, "learning_rate": 5.753550710142028e-06, "loss": 0.1113, "step": 10625 }, { "epoch": 3.3, "learning_rate": 5.753150630126025e-06, "loss": 0.0803, "step": 10626 }, { "epoch": 3.3, "learning_rate": 5.7527505501100226e-06, "loss": 0.0474, "step": 10627 }, { "epoch": 3.3, "learning_rate": 5.752350470094019e-06, "loss": 0.1272, "step": 10628 }, { "epoch": 3.3, "learning_rate": 5.751950390078016e-06, "loss": 0.0459, "step": 10629 }, { "epoch": 3.3, "learning_rate": 5.751550310062013e-06, "loss": 0.0337, "step": 10630 }, { "epoch": 3.3, "learning_rate": 5.75115023004601e-06, "loss": 0.044, "step": 10631 }, { "epoch": 3.3, "learning_rate": 5.750750150030007e-06, "loss": 0.0184, "step": 10632 }, { "epoch": 3.3, "learning_rate": 5.7503500700140035e-06, "loss": 0.0872, "step": 10633 }, { "epoch": 3.3, "learning_rate": 5.749949989998e-06, "loss": 0.0331, "step": 10634 }, { "epoch": 3.3, "learning_rate": 5.749549909981998e-06, "loss": 0.0499, "step": 10635 }, { "epoch": 3.3, "learning_rate": 5.749149829965993e-06, "loss": 0.0393, "step": 10636 }, { "epoch": 3.3, "learning_rate": 5.74874974994999e-06, "loss": 0.0328, "step": 10637 }, { "epoch": 3.3, "learning_rate": 5.748349669933987e-06, "loss": 0.0142, "step": 10638 }, { "epoch": 3.3, "learning_rate": 5.747949589917984e-06, "loss": 0.068, "step": 10639 }, { "epoch": 3.3, "learning_rate": 5.74754950990198e-06, "loss": 0.0305, "step": 10640 }, { "epoch": 3.3, "learning_rate": 5.747149429885978e-06, "loss": 0.0673, "step": 10641 }, { "epoch": 3.3, "learning_rate": 5.7467493498699746e-06, "loss": 0.0335, "step": 10642 }, { "epoch": 3.3, "learning_rate": 5.746349269853971e-06, "loss": 0.0249, "step": 10643 }, { "epoch": 3.3, "learning_rate": 5.745949189837968e-06, "loss": 0.0578, "step": 10644 }, { "epoch": 3.3, "learning_rate": 5.7455491098219655e-06, "loss": 0.0196, "step": 10645 }, { "epoch": 3.3, "learning_rate": 5.745149029805962e-06, "loss": 0.0317, "step": 10646 }, { "epoch": 3.3, "learning_rate": 5.744748949789959e-06, "loss": 0.0646, "step": 10647 }, { "epoch": 3.3, "learning_rate": 5.7443488697739555e-06, "loss": 0.0833, "step": 10648 }, { "epoch": 3.31, "learning_rate": 5.743948789757951e-06, "loss": 0.03, "step": 10649 }, { "epoch": 3.31, "learning_rate": 5.743548709741948e-06, "loss": 0.0553, "step": 10650 }, { "epoch": 3.31, "learning_rate": 5.743148629725946e-06, "loss": 0.0345, "step": 10651 }, { "epoch": 3.31, "learning_rate": 5.742748549709942e-06, "loss": 0.0261, "step": 10652 }, { "epoch": 3.31, "learning_rate": 5.742348469693939e-06, "loss": 0.0393, "step": 10653 }, { "epoch": 3.31, "learning_rate": 5.741948389677936e-06, "loss": 0.0263, "step": 10654 }, { "epoch": 3.31, "learning_rate": 5.741548309661933e-06, "loss": 0.0507, "step": 10655 }, { "epoch": 3.31, "learning_rate": 5.74114822964593e-06, "loss": 0.0369, "step": 10656 }, { "epoch": 3.31, "learning_rate": 5.7407481496299266e-06, "loss": 0.0399, "step": 10657 }, { "epoch": 3.31, "learning_rate": 5.740348069613923e-06, "loss": 0.1153, "step": 10658 }, { "epoch": 3.31, "learning_rate": 5.739947989597921e-06, "loss": 0.0347, "step": 10659 }, { "epoch": 3.31, "learning_rate": 5.7395479095819175e-06, "loss": 0.0287, "step": 10660 }, { "epoch": 3.31, "learning_rate": 5.739147829565913e-06, "loss": 0.0334, "step": 10661 }, { "epoch": 3.31, "learning_rate": 5.73874774954991e-06, "loss": 0.0775, "step": 10662 }, { "epoch": 3.31, "learning_rate": 5.738347669533907e-06, "loss": 0.0162, "step": 10663 }, { "epoch": 3.31, "learning_rate": 5.737947589517903e-06, "loss": 0.0369, "step": 10664 }, { "epoch": 3.31, "learning_rate": 5.737547509501901e-06, "loss": 0.0241, "step": 10665 }, { "epoch": 3.31, "learning_rate": 5.7371474294858976e-06, "loss": 0.0415, "step": 10666 }, { "epoch": 3.31, "learning_rate": 5.736747349469894e-06, "loss": 0.0743, "step": 10667 }, { "epoch": 3.31, "learning_rate": 5.736347269453891e-06, "loss": 0.1282, "step": 10668 }, { "epoch": 3.31, "learning_rate": 5.7359471894378885e-06, "loss": 0.0499, "step": 10669 }, { "epoch": 3.31, "learning_rate": 5.735547109421885e-06, "loss": 0.1614, "step": 10670 }, { "epoch": 3.31, "learning_rate": 5.735147029405882e-06, "loss": 0.0318, "step": 10671 }, { "epoch": 3.31, "learning_rate": 5.7347469493898785e-06, "loss": 0.0573, "step": 10672 }, { "epoch": 3.31, "learning_rate": 5.734346869373876e-06, "loss": 0.0419, "step": 10673 }, { "epoch": 3.31, "learning_rate": 5.733946789357872e-06, "loss": 0.0326, "step": 10674 }, { "epoch": 3.31, "learning_rate": 5.733546709341869e-06, "loss": 0.0271, "step": 10675 }, { "epoch": 3.31, "learning_rate": 5.733146629325865e-06, "loss": 0.0319, "step": 10676 }, { "epoch": 3.31, "learning_rate": 5.732746549309862e-06, "loss": 0.0604, "step": 10677 }, { "epoch": 3.31, "learning_rate": 5.732346469293859e-06, "loss": 0.021, "step": 10678 }, { "epoch": 3.31, "learning_rate": 5.731946389277856e-06, "loss": 0.1614, "step": 10679 }, { "epoch": 3.31, "learning_rate": 5.731546309261853e-06, "loss": 0.0368, "step": 10680 }, { "epoch": 3.32, "learning_rate": 5.7311462292458496e-06, "loss": 0.0402, "step": 10681 }, { "epoch": 3.32, "learning_rate": 5.730746149229846e-06, "loss": 0.0549, "step": 10682 }, { "epoch": 3.32, "learning_rate": 5.730346069213844e-06, "loss": 0.05, "step": 10683 }, { "epoch": 3.32, "learning_rate": 5.7299459891978405e-06, "loss": 0.0629, "step": 10684 }, { "epoch": 3.32, "learning_rate": 5.729545909181837e-06, "loss": 0.0269, "step": 10685 }, { "epoch": 3.32, "learning_rate": 5.729145829165834e-06, "loss": 0.0441, "step": 10686 }, { "epoch": 3.32, "learning_rate": 5.72874574914983e-06, "loss": 0.0424, "step": 10687 }, { "epoch": 3.32, "learning_rate": 5.728345669133827e-06, "loss": 0.039, "step": 10688 }, { "epoch": 3.32, "learning_rate": 5.727945589117824e-06, "loss": 0.0316, "step": 10689 }, { "epoch": 3.32, "learning_rate": 5.727545509101821e-06, "loss": 0.0123, "step": 10690 }, { "epoch": 3.32, "learning_rate": 5.727145429085817e-06, "loss": 0.0435, "step": 10691 }, { "epoch": 3.32, "learning_rate": 5.726745349069815e-06, "loss": 0.016, "step": 10692 }, { "epoch": 3.32, "learning_rate": 5.7263452690538115e-06, "loss": 0.0374, "step": 10693 }, { "epoch": 3.32, "learning_rate": 5.725945189037808e-06, "loss": 0.0414, "step": 10694 }, { "epoch": 3.32, "learning_rate": 5.725545109021805e-06, "loss": 0.05, "step": 10695 }, { "epoch": 3.32, "learning_rate": 5.7251450290058016e-06, "loss": 0.1247, "step": 10696 }, { "epoch": 3.32, "learning_rate": 5.724744948989799e-06, "loss": 0.0888, "step": 10697 }, { "epoch": 3.32, "learning_rate": 5.724344868973796e-06, "loss": 0.0233, "step": 10698 }, { "epoch": 3.32, "learning_rate": 5.723944788957792e-06, "loss": 0.0871, "step": 10699 }, { "epoch": 3.32, "learning_rate": 5.723544708941788e-06, "loss": 0.032, "step": 10700 }, { "epoch": 3.32, "learning_rate": 5.723144628925785e-06, "loss": 0.0576, "step": 10701 }, { "epoch": 3.32, "learning_rate": 5.7227445489097825e-06, "loss": 0.0174, "step": 10702 }, { "epoch": 3.32, "learning_rate": 5.722344468893779e-06, "loss": 0.0428, "step": 10703 }, { "epoch": 3.32, "learning_rate": 5.721944388877776e-06, "loss": 0.0359, "step": 10704 }, { "epoch": 3.32, "learning_rate": 5.7215443088617726e-06, "loss": 0.0223, "step": 10705 }, { "epoch": 3.32, "learning_rate": 5.72114422884577e-06, "loss": 0.0504, "step": 10706 }, { "epoch": 3.32, "learning_rate": 5.720744148829767e-06, "loss": 0.0546, "step": 10707 }, { "epoch": 3.32, "learning_rate": 5.7203440688137635e-06, "loss": 0.0351, "step": 10708 }, { "epoch": 3.32, "learning_rate": 5.71994398879776e-06, "loss": 0.0428, "step": 10709 }, { "epoch": 3.32, "learning_rate": 5.719543908781757e-06, "loss": 0.0283, "step": 10710 }, { "epoch": 3.32, "learning_rate": 5.719143828765754e-06, "loss": 0.094, "step": 10711 }, { "epoch": 3.32, "learning_rate": 5.71874374874975e-06, "loss": 0.0594, "step": 10712 }, { "epoch": 3.33, "learning_rate": 5.718343668733747e-06, "loss": 0.1123, "step": 10713 }, { "epoch": 3.33, "learning_rate": 5.717943588717744e-06, "loss": 0.0114, "step": 10714 }, { "epoch": 3.33, "learning_rate": 5.71754350870174e-06, "loss": 0.0372, "step": 10715 }, { "epoch": 3.33, "learning_rate": 5.717143428685738e-06, "loss": 0.0752, "step": 10716 }, { "epoch": 3.33, "learning_rate": 5.7167433486697345e-06, "loss": 0.025, "step": 10717 }, { "epoch": 3.33, "learning_rate": 5.716343268653731e-06, "loss": 0.0752, "step": 10718 }, { "epoch": 3.33, "learning_rate": 5.715943188637728e-06, "loss": 0.0424, "step": 10719 }, { "epoch": 3.33, "learning_rate": 5.715543108621725e-06, "loss": 0.0396, "step": 10720 }, { "epoch": 3.33, "learning_rate": 5.715143028605722e-06, "loss": 0.076, "step": 10721 }, { "epoch": 3.33, "learning_rate": 5.714742948589719e-06, "loss": 0.0439, "step": 10722 }, { "epoch": 3.33, "learning_rate": 5.7143428685737155e-06, "loss": 0.1974, "step": 10723 }, { "epoch": 3.33, "learning_rate": 5.713942788557712e-06, "loss": 0.0877, "step": 10724 }, { "epoch": 3.33, "learning_rate": 5.713542708541708e-06, "loss": 0.0435, "step": 10725 }, { "epoch": 3.33, "learning_rate": 5.7131426285257055e-06, "loss": 0.032, "step": 10726 }, { "epoch": 3.33, "learning_rate": 5.712742548509702e-06, "loss": 0.0185, "step": 10727 }, { "epoch": 3.33, "learning_rate": 5.712342468493699e-06, "loss": 0.0339, "step": 10728 }, { "epoch": 3.33, "learning_rate": 5.711942388477696e-06, "loss": 0.0249, "step": 10729 }, { "epoch": 3.33, "learning_rate": 5.711542308461693e-06, "loss": 0.0431, "step": 10730 }, { "epoch": 3.33, "learning_rate": 5.71114222844569e-06, "loss": 0.0271, "step": 10731 }, { "epoch": 3.33, "learning_rate": 5.7107421484296865e-06, "loss": 0.0468, "step": 10732 }, { "epoch": 3.33, "learning_rate": 5.710342068413683e-06, "loss": 0.0352, "step": 10733 }, { "epoch": 3.33, "learning_rate": 5.709941988397681e-06, "loss": 0.0196, "step": 10734 }, { "epoch": 3.33, "learning_rate": 5.709541908381677e-06, "loss": 0.0293, "step": 10735 }, { "epoch": 3.33, "learning_rate": 5.709141828365674e-06, "loss": 0.0399, "step": 10736 }, { "epoch": 3.33, "learning_rate": 5.70874174834967e-06, "loss": 0.0739, "step": 10737 }, { "epoch": 3.33, "learning_rate": 5.708341668333667e-06, "loss": 0.0148, "step": 10738 }, { "epoch": 3.33, "learning_rate": 5.707941588317663e-06, "loss": 0.0452, "step": 10739 }, { "epoch": 3.33, "learning_rate": 5.707541508301661e-06, "loss": 0.0137, "step": 10740 }, { "epoch": 3.33, "learning_rate": 5.7071414282856575e-06, "loss": 0.0159, "step": 10741 }, { "epoch": 3.33, "learning_rate": 5.706741348269654e-06, "loss": 0.0234, "step": 10742 }, { "epoch": 3.33, "learning_rate": 5.706341268253651e-06, "loss": 0.0974, "step": 10743 }, { "epoch": 3.33, "learning_rate": 5.705941188237648e-06, "loss": 0.1586, "step": 10744 }, { "epoch": 3.34, "learning_rate": 5.705541108221645e-06, "loss": 0.0192, "step": 10745 }, { "epoch": 3.34, "learning_rate": 5.705141028205642e-06, "loss": 0.0396, "step": 10746 }, { "epoch": 3.34, "learning_rate": 5.7047409481896385e-06, "loss": 0.0489, "step": 10747 }, { "epoch": 3.34, "learning_rate": 5.704340868173636e-06, "loss": 0.0915, "step": 10748 }, { "epoch": 3.34, "learning_rate": 5.703940788157633e-06, "loss": 0.0604, "step": 10749 }, { "epoch": 3.34, "learning_rate": 5.7035407081416285e-06, "loss": 0.0551, "step": 10750 }, { "epoch": 3.34, "learning_rate": 5.703140628125625e-06, "loss": 0.0336, "step": 10751 }, { "epoch": 3.34, "learning_rate": 5.702740548109622e-06, "loss": 0.0296, "step": 10752 }, { "epoch": 3.34, "learning_rate": 5.702340468093619e-06, "loss": 0.0472, "step": 10753 }, { "epoch": 3.34, "learning_rate": 5.701940388077616e-06, "loss": 0.0685, "step": 10754 }, { "epoch": 3.34, "learning_rate": 5.701540308061613e-06, "loss": 0.0322, "step": 10755 }, { "epoch": 3.34, "learning_rate": 5.7011402280456095e-06, "loss": 0.1352, "step": 10756 }, { "epoch": 3.34, "learning_rate": 5.700740148029606e-06, "loss": 0.0706, "step": 10757 }, { "epoch": 3.34, "learning_rate": 5.700340068013604e-06, "loss": 0.0321, "step": 10758 }, { "epoch": 3.34, "learning_rate": 5.6999399879976e-06, "loss": 0.0515, "step": 10759 }, { "epoch": 3.34, "learning_rate": 5.699539907981597e-06, "loss": 0.0636, "step": 10760 }, { "epoch": 3.34, "learning_rate": 5.699139827965594e-06, "loss": 0.0532, "step": 10761 }, { "epoch": 3.34, "learning_rate": 5.69873974794959e-06, "loss": 0.0757, "step": 10762 }, { "epoch": 3.34, "learning_rate": 5.698339667933586e-06, "loss": 0.1562, "step": 10763 }, { "epoch": 3.34, "learning_rate": 5.697939587917584e-06, "loss": 0.0562, "step": 10764 }, { "epoch": 3.34, "learning_rate": 5.6975395079015805e-06, "loss": 0.0187, "step": 10765 }, { "epoch": 3.34, "learning_rate": 5.697139427885577e-06, "loss": 0.0185, "step": 10766 }, { "epoch": 3.34, "learning_rate": 5.696739347869574e-06, "loss": 0.0376, "step": 10767 }, { "epoch": 3.34, "learning_rate": 5.6963392678535714e-06, "loss": 0.0425, "step": 10768 }, { "epoch": 3.34, "learning_rate": 5.695939187837568e-06, "loss": 0.039, "step": 10769 }, { "epoch": 3.34, "learning_rate": 5.695539107821565e-06, "loss": 0.0353, "step": 10770 }, { "epoch": 3.34, "learning_rate": 5.6951390278055615e-06, "loss": 0.0293, "step": 10771 }, { "epoch": 3.34, "learning_rate": 5.694738947789559e-06, "loss": 0.0234, "step": 10772 }, { "epoch": 3.34, "learning_rate": 5.694338867773556e-06, "loss": 0.0645, "step": 10773 }, { "epoch": 3.34, "learning_rate": 5.693938787757552e-06, "loss": 0.0195, "step": 10774 }, { "epoch": 3.34, "learning_rate": 5.693538707741548e-06, "loss": 0.054, "step": 10775 }, { "epoch": 3.34, "learning_rate": 5.693138627725545e-06, "loss": 0.0366, "step": 10776 }, { "epoch": 3.34, "learning_rate": 5.6927385477095425e-06, "loss": 0.0337, "step": 10777 }, { "epoch": 3.35, "learning_rate": 5.692338467693539e-06, "loss": 0.1154, "step": 10778 }, { "epoch": 3.35, "learning_rate": 5.691938387677536e-06, "loss": 0.196, "step": 10779 }, { "epoch": 3.35, "learning_rate": 5.6915383076615325e-06, "loss": 0.0577, "step": 10780 }, { "epoch": 3.35, "learning_rate": 5.691138227645529e-06, "loss": 0.0441, "step": 10781 }, { "epoch": 3.35, "learning_rate": 5.690738147629527e-06, "loss": 0.032, "step": 10782 }, { "epoch": 3.35, "learning_rate": 5.690338067613523e-06, "loss": 0.0339, "step": 10783 }, { "epoch": 3.35, "learning_rate": 5.68993798759752e-06, "loss": 0.0446, "step": 10784 }, { "epoch": 3.35, "learning_rate": 5.689537907581517e-06, "loss": 0.0353, "step": 10785 }, { "epoch": 3.35, "learning_rate": 5.689137827565514e-06, "loss": 0.0297, "step": 10786 }, { "epoch": 3.35, "learning_rate": 5.688737747549511e-06, "loss": 0.052, "step": 10787 }, { "epoch": 3.35, "learning_rate": 5.688337667533507e-06, "loss": 0.0307, "step": 10788 }, { "epoch": 3.35, "learning_rate": 5.6879375875175035e-06, "loss": 0.0373, "step": 10789 }, { "epoch": 3.35, "learning_rate": 5.6875375075015e-06, "loss": 0.0426, "step": 10790 }, { "epoch": 3.35, "learning_rate": 5.687137427485498e-06, "loss": 0.0327, "step": 10791 }, { "epoch": 3.35, "learning_rate": 5.6867373474694944e-06, "loss": 0.032, "step": 10792 }, { "epoch": 3.35, "learning_rate": 5.686337267453491e-06, "loss": 0.0357, "step": 10793 }, { "epoch": 3.35, "learning_rate": 5.685937187437488e-06, "loss": 0.0317, "step": 10794 }, { "epoch": 3.35, "learning_rate": 5.6855371074214845e-06, "loss": 0.0345, "step": 10795 }, { "epoch": 3.35, "learning_rate": 5.685137027405482e-06, "loss": 0.1307, "step": 10796 }, { "epoch": 3.35, "learning_rate": 5.684736947389479e-06, "loss": 0.0429, "step": 10797 }, { "epoch": 3.35, "learning_rate": 5.684336867373475e-06, "loss": 0.0454, "step": 10798 }, { "epoch": 3.35, "learning_rate": 5.683936787357472e-06, "loss": 0.0426, "step": 10799 }, { "epoch": 3.35, "learning_rate": 5.683536707341468e-06, "loss": 0.0434, "step": 10800 }, { "epoch": 3.35, "learning_rate": 5.6831366273254655e-06, "loss": 0.1242, "step": 10801 }, { "epoch": 3.35, "learning_rate": 5.682736547309462e-06, "loss": 0.052, "step": 10802 }, { "epoch": 3.35, "learning_rate": 5.682336467293459e-06, "loss": 0.037, "step": 10803 }, { "epoch": 3.35, "learning_rate": 5.6819363872774555e-06, "loss": 0.023, "step": 10804 }, { "epoch": 3.35, "learning_rate": 5.681536307261453e-06, "loss": 0.0573, "step": 10805 }, { "epoch": 3.35, "learning_rate": 5.68113622724545e-06, "loss": 0.0377, "step": 10806 }, { "epoch": 3.35, "learning_rate": 5.6807361472294464e-06, "loss": 0.0542, "step": 10807 }, { "epoch": 3.35, "learning_rate": 5.680336067213443e-06, "loss": 0.0352, "step": 10808 }, { "epoch": 3.35, "learning_rate": 5.67993598719744e-06, "loss": 0.0432, "step": 10809 }, { "epoch": 3.36, "learning_rate": 5.679535907181437e-06, "loss": 0.0454, "step": 10810 }, { "epoch": 3.36, "learning_rate": 5.679135827165434e-06, "loss": 0.0581, "step": 10811 }, { "epoch": 3.36, "learning_rate": 5.678735747149431e-06, "loss": 0.0591, "step": 10812 }, { "epoch": 3.36, "learning_rate": 5.6783356671334266e-06, "loss": 0.0668, "step": 10813 }, { "epoch": 3.36, "learning_rate": 5.677935587117423e-06, "loss": 0.0699, "step": 10814 }, { "epoch": 3.36, "learning_rate": 5.677535507101421e-06, "loss": 0.0208, "step": 10815 }, { "epoch": 3.36, "learning_rate": 5.6771354270854175e-06, "loss": 0.0465, "step": 10816 }, { "epoch": 3.36, "learning_rate": 5.676735347069414e-06, "loss": 0.0387, "step": 10817 }, { "epoch": 3.36, "learning_rate": 5.676335267053411e-06, "loss": 0.0267, "step": 10818 }, { "epoch": 3.36, "learning_rate": 5.675935187037408e-06, "loss": 0.0611, "step": 10819 }, { "epoch": 3.36, "learning_rate": 5.675535107021405e-06, "loss": 0.0655, "step": 10820 }, { "epoch": 3.36, "learning_rate": 5.675135027005402e-06, "loss": 0.0739, "step": 10821 }, { "epoch": 3.36, "learning_rate": 5.674734946989398e-06, "loss": 0.031, "step": 10822 }, { "epoch": 3.36, "learning_rate": 5.674334866973395e-06, "loss": 0.0447, "step": 10823 }, { "epoch": 3.36, "learning_rate": 5.673934786957393e-06, "loss": 0.0373, "step": 10824 }, { "epoch": 3.36, "learning_rate": 5.6735347069413885e-06, "loss": 0.0665, "step": 10825 }, { "epoch": 3.36, "learning_rate": 5.673134626925385e-06, "loss": 0.0246, "step": 10826 }, { "epoch": 3.36, "learning_rate": 5.672734546909382e-06, "loss": 0.0699, "step": 10827 }, { "epoch": 3.36, "learning_rate": 5.6723344668933785e-06, "loss": 0.0252, "step": 10828 }, { "epoch": 3.36, "learning_rate": 5.671934386877376e-06, "loss": 0.0479, "step": 10829 }, { "epoch": 3.36, "learning_rate": 5.671534306861373e-06, "loss": 0.0169, "step": 10830 }, { "epoch": 3.36, "learning_rate": 5.6711342268453694e-06, "loss": 0.0571, "step": 10831 }, { "epoch": 3.36, "learning_rate": 5.670734146829366e-06, "loss": 0.0618, "step": 10832 }, { "epoch": 3.36, "learning_rate": 5.670334066813364e-06, "loss": 0.0191, "step": 10833 }, { "epoch": 3.36, "learning_rate": 5.66993398679736e-06, "loss": 0.0198, "step": 10834 }, { "epoch": 3.36, "learning_rate": 5.669533906781357e-06, "loss": 0.0507, "step": 10835 }, { "epoch": 3.36, "learning_rate": 5.669133826765354e-06, "loss": 0.0415, "step": 10836 }, { "epoch": 3.36, "learning_rate": 5.668733746749351e-06, "loss": 0.0304, "step": 10837 }, { "epoch": 3.36, "learning_rate": 5.668333666733346e-06, "loss": 0.0202, "step": 10838 }, { "epoch": 3.36, "learning_rate": 5.667933586717344e-06, "loss": 0.0527, "step": 10839 }, { "epoch": 3.36, "learning_rate": 5.6675335067013405e-06, "loss": 0.0723, "step": 10840 }, { "epoch": 3.36, "learning_rate": 5.667133426685337e-06, "loss": 0.0242, "step": 10841 }, { "epoch": 3.37, "learning_rate": 5.666733346669334e-06, "loss": 0.0495, "step": 10842 }, { "epoch": 3.37, "learning_rate": 5.666333266653331e-06, "loss": 0.0475, "step": 10843 }, { "epoch": 3.37, "learning_rate": 5.665933186637328e-06, "loss": 0.0427, "step": 10844 }, { "epoch": 3.37, "learning_rate": 5.665533106621325e-06, "loss": 0.0297, "step": 10845 }, { "epoch": 3.37, "learning_rate": 5.6651330266053214e-06, "loss": 0.0215, "step": 10846 }, { "epoch": 3.37, "learning_rate": 5.664732946589319e-06, "loss": 0.0373, "step": 10847 }, { "epoch": 3.37, "learning_rate": 5.664332866573316e-06, "loss": 0.0757, "step": 10848 }, { "epoch": 3.37, "learning_rate": 5.663932786557312e-06, "loss": 0.0417, "step": 10849 }, { "epoch": 3.37, "learning_rate": 5.663532706541309e-06, "loss": 0.057, "step": 10850 }, { "epoch": 3.37, "learning_rate": 5.663132626525305e-06, "loss": 0.0581, "step": 10851 }, { "epoch": 3.37, "learning_rate": 5.6627325465093016e-06, "loss": 0.0633, "step": 10852 }, { "epoch": 3.37, "learning_rate": 5.662332466493299e-06, "loss": 0.1237, "step": 10853 }, { "epoch": 3.37, "learning_rate": 5.661932386477296e-06, "loss": 0.0412, "step": 10854 }, { "epoch": 3.37, "learning_rate": 5.6615323064612925e-06, "loss": 0.0174, "step": 10855 }, { "epoch": 3.37, "learning_rate": 5.661132226445289e-06, "loss": 0.0532, "step": 10856 }, { "epoch": 3.37, "learning_rate": 5.660732146429287e-06, "loss": 0.058, "step": 10857 }, { "epoch": 3.37, "learning_rate": 5.660332066413283e-06, "loss": 0.0472, "step": 10858 }, { "epoch": 3.37, "learning_rate": 5.65993198639728e-06, "loss": 0.0499, "step": 10859 }, { "epoch": 3.37, "learning_rate": 5.659531906381277e-06, "loss": 0.0632, "step": 10860 }, { "epoch": 3.37, "learning_rate": 5.659131826365274e-06, "loss": 0.0409, "step": 10861 }, { "epoch": 3.37, "learning_rate": 5.658731746349271e-06, "loss": 0.0321, "step": 10862 }, { "epoch": 3.37, "learning_rate": 5.658331666333267e-06, "loss": 0.0156, "step": 10863 }, { "epoch": 3.37, "learning_rate": 5.6579315863172635e-06, "loss": 0.044, "step": 10864 }, { "epoch": 3.37, "learning_rate": 5.65753150630126e-06, "loss": 0.0339, "step": 10865 }, { "epoch": 3.37, "learning_rate": 5.657131426285257e-06, "loss": 0.0734, "step": 10866 }, { "epoch": 3.37, "learning_rate": 5.656731346269254e-06, "loss": 0.0164, "step": 10867 }, { "epoch": 3.37, "learning_rate": 5.656331266253251e-06, "loss": 0.0395, "step": 10868 }, { "epoch": 3.37, "learning_rate": 5.655931186237248e-06, "loss": 0.0212, "step": 10869 }, { "epoch": 3.37, "learning_rate": 5.6555311062212444e-06, "loss": 0.0152, "step": 10870 }, { "epoch": 3.37, "learning_rate": 5.655131026205242e-06, "loss": 0.0134, "step": 10871 }, { "epoch": 3.37, "learning_rate": 5.654730946189239e-06, "loss": 0.1154, "step": 10872 }, { "epoch": 3.37, "learning_rate": 5.654330866173235e-06, "loss": 0.1082, "step": 10873 }, { "epoch": 3.38, "learning_rate": 5.653930786157232e-06, "loss": 0.0414, "step": 10874 }, { "epoch": 3.38, "learning_rate": 5.6535307061412296e-06, "loss": 0.0138, "step": 10875 }, { "epoch": 3.38, "learning_rate": 5.653130626125225e-06, "loss": 0.0439, "step": 10876 }, { "epoch": 3.38, "learning_rate": 5.652730546109222e-06, "loss": 0.0263, "step": 10877 }, { "epoch": 3.38, "learning_rate": 5.652330466093219e-06, "loss": 0.0256, "step": 10878 }, { "epoch": 3.38, "learning_rate": 5.6519303860772155e-06, "loss": 0.096, "step": 10879 }, { "epoch": 3.38, "learning_rate": 5.651530306061212e-06, "loss": 0.0538, "step": 10880 }, { "epoch": 3.38, "learning_rate": 5.65113022604521e-06, "loss": 0.0555, "step": 10881 }, { "epoch": 3.38, "learning_rate": 5.650730146029206e-06, "loss": 0.0306, "step": 10882 }, { "epoch": 3.38, "learning_rate": 5.650330066013203e-06, "loss": 0.0608, "step": 10883 }, { "epoch": 3.38, "learning_rate": 5.6499299859972e-06, "loss": 0.0406, "step": 10884 }, { "epoch": 3.38, "learning_rate": 5.649529905981197e-06, "loss": 0.037, "step": 10885 }, { "epoch": 3.38, "learning_rate": 5.649129825965194e-06, "loss": 0.0179, "step": 10886 }, { "epoch": 3.38, "learning_rate": 5.648729745949191e-06, "loss": 0.0594, "step": 10887 }, { "epoch": 3.38, "learning_rate": 5.6483296659331865e-06, "loss": 0.1433, "step": 10888 }, { "epoch": 3.38, "learning_rate": 5.647929585917183e-06, "loss": 0.0319, "step": 10889 }, { "epoch": 3.38, "learning_rate": 5.647529505901181e-06, "loss": 0.0359, "step": 10890 }, { "epoch": 3.38, "learning_rate": 5.647129425885177e-06, "loss": 0.0666, "step": 10891 }, { "epoch": 3.38, "learning_rate": 5.646729345869174e-06, "loss": 0.0315, "step": 10892 }, { "epoch": 3.38, "learning_rate": 5.646329265853171e-06, "loss": 0.0691, "step": 10893 }, { "epoch": 3.38, "learning_rate": 5.6459291858371675e-06, "loss": 0.0407, "step": 10894 }, { "epoch": 3.38, "learning_rate": 5.645529105821165e-06, "loss": 0.0553, "step": 10895 }, { "epoch": 3.38, "learning_rate": 5.645129025805162e-06, "loss": 0.0752, "step": 10896 }, { "epoch": 3.38, "learning_rate": 5.644728945789158e-06, "loss": 0.043, "step": 10897 }, { "epoch": 3.38, "learning_rate": 5.644328865773155e-06, "loss": 0.0236, "step": 10898 }, { "epoch": 3.38, "learning_rate": 5.643928785757153e-06, "loss": 0.0484, "step": 10899 }, { "epoch": 3.38, "learning_rate": 5.643528705741149e-06, "loss": 0.0401, "step": 10900 }, { "epoch": 3.38, "learning_rate": 5.643128625725145e-06, "loss": 0.0238, "step": 10901 }, { "epoch": 3.38, "learning_rate": 5.642728545709142e-06, "loss": 0.0208, "step": 10902 }, { "epoch": 3.38, "learning_rate": 5.6423284656931385e-06, "loss": 0.0337, "step": 10903 }, { "epoch": 3.38, "learning_rate": 5.641928385677136e-06, "loss": 0.0678, "step": 10904 }, { "epoch": 3.38, "learning_rate": 5.641528305661133e-06, "loss": 0.0826, "step": 10905 }, { "epoch": 3.38, "learning_rate": 5.641128225645129e-06, "loss": 0.0187, "step": 10906 }, { "epoch": 3.39, "learning_rate": 5.640728145629126e-06, "loss": 0.036, "step": 10907 }, { "epoch": 3.39, "learning_rate": 5.640328065613124e-06, "loss": 0.0427, "step": 10908 }, { "epoch": 3.39, "learning_rate": 5.63992798559712e-06, "loss": 0.0468, "step": 10909 }, { "epoch": 3.39, "learning_rate": 5.639527905581117e-06, "loss": 0.0186, "step": 10910 }, { "epoch": 3.39, "learning_rate": 5.639127825565114e-06, "loss": 0.0462, "step": 10911 }, { "epoch": 3.39, "learning_rate": 5.63872774554911e-06, "loss": 0.022, "step": 10912 }, { "epoch": 3.39, "learning_rate": 5.638327665533108e-06, "loss": 0.0527, "step": 10913 }, { "epoch": 3.39, "learning_rate": 5.637927585517104e-06, "loss": 0.0296, "step": 10914 }, { "epoch": 3.39, "learning_rate": 5.6375275055011e-06, "loss": 0.0981, "step": 10915 }, { "epoch": 3.39, "learning_rate": 5.637127425485097e-06, "loss": 0.0308, "step": 10916 }, { "epoch": 3.39, "learning_rate": 5.636727345469094e-06, "loss": 0.0366, "step": 10917 }, { "epoch": 3.39, "learning_rate": 5.636327265453091e-06, "loss": 0.071, "step": 10918 }, { "epoch": 3.39, "learning_rate": 5.635927185437088e-06, "loss": 0.0239, "step": 10919 }, { "epoch": 3.39, "learning_rate": 5.635527105421085e-06, "loss": 0.0303, "step": 10920 }, { "epoch": 3.39, "learning_rate": 5.635127025405081e-06, "loss": 0.0453, "step": 10921 }, { "epoch": 3.39, "learning_rate": 5.634726945389079e-06, "loss": 0.0535, "step": 10922 }, { "epoch": 3.39, "learning_rate": 5.634326865373076e-06, "loss": 0.0111, "step": 10923 }, { "epoch": 3.39, "learning_rate": 5.633926785357072e-06, "loss": 0.0265, "step": 10924 }, { "epoch": 3.39, "learning_rate": 5.633526705341069e-06, "loss": 0.0522, "step": 10925 }, { "epoch": 3.39, "learning_rate": 5.633126625325065e-06, "loss": 0.0162, "step": 10926 }, { "epoch": 3.39, "learning_rate": 5.6327265453090615e-06, "loss": 0.0332, "step": 10927 }, { "epoch": 3.39, "learning_rate": 5.632326465293059e-06, "loss": 0.0365, "step": 10928 }, { "epoch": 3.39, "learning_rate": 5.631926385277056e-06, "loss": 0.0448, "step": 10929 }, { "epoch": 3.39, "learning_rate": 5.631526305261052e-06, "loss": 0.0208, "step": 10930 }, { "epoch": 3.39, "learning_rate": 5.631126225245049e-06, "loss": 0.0288, "step": 10931 }, { "epoch": 3.39, "learning_rate": 5.630726145229047e-06, "loss": 0.0809, "step": 10932 }, { "epoch": 3.39, "learning_rate": 5.630326065213043e-06, "loss": 0.0316, "step": 10933 }, { "epoch": 3.39, "learning_rate": 5.62992598519704e-06, "loss": 0.0223, "step": 10934 }, { "epoch": 3.39, "learning_rate": 5.629525905181037e-06, "loss": 0.0295, "step": 10935 }, { "epoch": 3.39, "learning_rate": 5.629125825165034e-06, "loss": 0.0342, "step": 10936 }, { "epoch": 3.39, "learning_rate": 5.628725745149031e-06, "loss": 0.0712, "step": 10937 }, { "epoch": 3.39, "learning_rate": 5.628325665133028e-06, "loss": 0.0901, "step": 10938 }, { "epoch": 3.4, "learning_rate": 5.627925585117023e-06, "loss": 0.0735, "step": 10939 }, { "epoch": 3.4, "learning_rate": 5.62752550510102e-06, "loss": 0.0422, "step": 10940 }, { "epoch": 3.4, "learning_rate": 5.627125425085017e-06, "loss": 0.1432, "step": 10941 }, { "epoch": 3.4, "learning_rate": 5.626725345069014e-06, "loss": 0.1673, "step": 10942 }, { "epoch": 3.4, "learning_rate": 5.626325265053011e-06, "loss": 0.1445, "step": 10943 }, { "epoch": 3.4, "learning_rate": 5.625925185037008e-06, "loss": 0.1208, "step": 10944 }, { "epoch": 3.4, "learning_rate": 5.625525105021004e-06, "loss": 0.0552, "step": 10945 }, { "epoch": 3.4, "learning_rate": 5.625125025005002e-06, "loss": 0.0386, "step": 10946 }, { "epoch": 3.4, "learning_rate": 5.624724944988999e-06, "loss": 0.0138, "step": 10947 }, { "epoch": 3.4, "learning_rate": 5.624324864972995e-06, "loss": 0.0719, "step": 10948 }, { "epoch": 3.4, "learning_rate": 5.623924784956992e-06, "loss": 0.0539, "step": 10949 }, { "epoch": 3.4, "learning_rate": 5.6235247049409895e-06, "loss": 0.0416, "step": 10950 }, { "epoch": 3.4, "learning_rate": 5.623124624924986e-06, "loss": 0.0235, "step": 10951 }, { "epoch": 3.4, "learning_rate": 5.622724544908982e-06, "loss": 0.1063, "step": 10952 }, { "epoch": 3.4, "learning_rate": 5.622324464892979e-06, "loss": 0.0637, "step": 10953 }, { "epoch": 3.4, "learning_rate": 5.621924384876975e-06, "loss": 0.0295, "step": 10954 }, { "epoch": 3.4, "learning_rate": 5.621524304860972e-06, "loss": 0.0244, "step": 10955 }, { "epoch": 3.4, "learning_rate": 5.62112422484497e-06, "loss": 0.0402, "step": 10956 }, { "epoch": 3.4, "learning_rate": 5.620724144828966e-06, "loss": 0.0294, "step": 10957 }, { "epoch": 3.4, "learning_rate": 5.620324064812963e-06, "loss": 0.0479, "step": 10958 }, { "epoch": 3.4, "learning_rate": 5.61992398479696e-06, "loss": 0.0489, "step": 10959 }, { "epoch": 3.4, "learning_rate": 5.619523904780957e-06, "loss": 0.0668, "step": 10960 }, { "epoch": 3.4, "learning_rate": 5.619123824764954e-06, "loss": 0.0136, "step": 10961 }, { "epoch": 3.4, "learning_rate": 5.618723744748951e-06, "loss": 0.0278, "step": 10962 }, { "epoch": 3.4, "learning_rate": 5.618323664732947e-06, "loss": 0.1303, "step": 10963 }, { "epoch": 3.4, "learning_rate": 5.617923584716943e-06, "loss": 0.0552, "step": 10964 }, { "epoch": 3.4, "learning_rate": 5.61752350470094e-06, "loss": 0.0207, "step": 10965 }, { "epoch": 3.4, "learning_rate": 5.617123424684937e-06, "loss": 0.022, "step": 10966 }, { "epoch": 3.4, "learning_rate": 5.616723344668934e-06, "loss": 0.0612, "step": 10967 }, { "epoch": 3.4, "learning_rate": 5.616323264652931e-06, "loss": 0.0311, "step": 10968 }, { "epoch": 3.4, "learning_rate": 5.615923184636927e-06, "loss": 0.0947, "step": 10969 }, { "epoch": 3.4, "learning_rate": 5.615523104620925e-06, "loss": 0.0264, "step": 10970 }, { "epoch": 3.41, "learning_rate": 5.615123024604922e-06, "loss": 0.024, "step": 10971 }, { "epoch": 3.41, "learning_rate": 5.614722944588918e-06, "loss": 0.0566, "step": 10972 }, { "epoch": 3.41, "learning_rate": 5.614322864572915e-06, "loss": 0.0893, "step": 10973 }, { "epoch": 3.41, "learning_rate": 5.6139227845569125e-06, "loss": 0.0722, "step": 10974 }, { "epoch": 3.41, "learning_rate": 5.613522704540909e-06, "loss": 0.0272, "step": 10975 }, { "epoch": 3.41, "learning_rate": 5.613122624524906e-06, "loss": 0.0289, "step": 10976 }, { "epoch": 3.41, "learning_rate": 5.612722544508902e-06, "loss": 0.033, "step": 10977 }, { "epoch": 3.41, "learning_rate": 5.612322464492898e-06, "loss": 0.026, "step": 10978 }, { "epoch": 3.41, "learning_rate": 5.611922384476895e-06, "loss": 0.1194, "step": 10979 }, { "epoch": 3.41, "learning_rate": 5.611522304460893e-06, "loss": 0.155, "step": 10980 }, { "epoch": 3.41, "learning_rate": 5.611122224444889e-06, "loss": 0.0196, "step": 10981 }, { "epoch": 3.41, "learning_rate": 5.610722144428886e-06, "loss": 0.0402, "step": 10982 }, { "epoch": 3.41, "learning_rate": 5.610322064412883e-06, "loss": 0.0418, "step": 10983 }, { "epoch": 3.41, "learning_rate": 5.60992198439688e-06, "loss": 0.0314, "step": 10984 }, { "epoch": 3.41, "learning_rate": 5.609521904380877e-06, "loss": 0.0167, "step": 10985 }, { "epoch": 3.41, "learning_rate": 5.609121824364874e-06, "loss": 0.0291, "step": 10986 }, { "epoch": 3.41, "learning_rate": 5.60872174434887e-06, "loss": 0.0467, "step": 10987 }, { "epoch": 3.41, "learning_rate": 5.608321664332868e-06, "loss": 0.0296, "step": 10988 }, { "epoch": 3.41, "learning_rate": 5.607921584316864e-06, "loss": 0.1132, "step": 10989 }, { "epoch": 3.41, "learning_rate": 5.60752150430086e-06, "loss": 0.1347, "step": 10990 }, { "epoch": 3.41, "learning_rate": 5.607121424284857e-06, "loss": 0.0556, "step": 10991 }, { "epoch": 3.41, "learning_rate": 5.606721344268854e-06, "loss": 0.0752, "step": 10992 }, { "epoch": 3.41, "learning_rate": 5.606321264252851e-06, "loss": 0.0246, "step": 10993 }, { "epoch": 3.41, "learning_rate": 5.605921184236848e-06, "loss": 0.0271, "step": 10994 }, { "epoch": 3.41, "learning_rate": 5.605521104220845e-06, "loss": 0.0395, "step": 10995 }, { "epoch": 3.41, "learning_rate": 5.605121024204841e-06, "loss": 0.1338, "step": 10996 }, { "epoch": 3.41, "learning_rate": 5.604720944188838e-06, "loss": 0.0833, "step": 10997 }, { "epoch": 3.41, "learning_rate": 5.6043208641728355e-06, "loss": 0.0594, "step": 10998 }, { "epoch": 3.41, "learning_rate": 5.603920784156832e-06, "loss": 0.0524, "step": 10999 }, { "epoch": 3.41, "learning_rate": 5.603520704140829e-06, "loss": 0.0417, "step": 11000 }, { "epoch": 3.41, "learning_rate": 5.603120624124826e-06, "loss": 0.0641, "step": 11001 }, { "epoch": 3.41, "learning_rate": 5.6027205441088214e-06, "loss": 0.0589, "step": 11002 }, { "epoch": 3.42, "learning_rate": 5.602320464092819e-06, "loss": 0.0273, "step": 11003 }, { "epoch": 3.42, "learning_rate": 5.601920384076816e-06, "loss": 0.0333, "step": 11004 }, { "epoch": 3.42, "learning_rate": 5.601520304060812e-06, "loss": 0.1241, "step": 11005 }, { "epoch": 3.42, "learning_rate": 5.601120224044809e-06, "loss": 0.05, "step": 11006 }, { "epoch": 3.42, "learning_rate": 5.6007201440288066e-06, "loss": 0.0964, "step": 11007 }, { "epoch": 3.42, "learning_rate": 5.600320064012803e-06, "loss": 0.0636, "step": 11008 }, { "epoch": 3.42, "learning_rate": 5.5999199839968e-06, "loss": 0.0399, "step": 11009 }, { "epoch": 3.42, "learning_rate": 5.599519903980797e-06, "loss": 0.0283, "step": 11010 }, { "epoch": 3.42, "learning_rate": 5.599119823964793e-06, "loss": 0.0611, "step": 11011 }, { "epoch": 3.42, "learning_rate": 5.598719743948791e-06, "loss": 0.0435, "step": 11012 }, { "epoch": 3.42, "learning_rate": 5.5983196639327875e-06, "loss": 0.0499, "step": 11013 }, { "epoch": 3.42, "learning_rate": 5.597919583916784e-06, "loss": 0.0404, "step": 11014 }, { "epoch": 3.42, "learning_rate": 5.59751950390078e-06, "loss": 0.0312, "step": 11015 }, { "epoch": 3.42, "learning_rate": 5.597119423884777e-06, "loss": 0.0394, "step": 11016 }, { "epoch": 3.42, "learning_rate": 5.596719343868774e-06, "loss": 0.029, "step": 11017 }, { "epoch": 3.42, "learning_rate": 5.596319263852771e-06, "loss": 0.0281, "step": 11018 }, { "epoch": 3.42, "learning_rate": 5.595919183836768e-06, "loss": 0.0364, "step": 11019 }, { "epoch": 3.42, "learning_rate": 5.595519103820764e-06, "loss": 0.0664, "step": 11020 }, { "epoch": 3.42, "learning_rate": 5.595119023804762e-06, "loss": 0.0489, "step": 11021 }, { "epoch": 3.42, "learning_rate": 5.5947189437887585e-06, "loss": 0.0594, "step": 11022 }, { "epoch": 3.42, "learning_rate": 5.594318863772755e-06, "loss": 0.011, "step": 11023 }, { "epoch": 3.42, "learning_rate": 5.593918783756752e-06, "loss": 0.0885, "step": 11024 }, { "epoch": 3.42, "learning_rate": 5.593518703740749e-06, "loss": 0.0407, "step": 11025 }, { "epoch": 3.42, "learning_rate": 5.593118623724746e-06, "loss": 0.0451, "step": 11026 }, { "epoch": 3.42, "learning_rate": 5.592718543708742e-06, "loss": 0.0327, "step": 11027 }, { "epoch": 3.42, "learning_rate": 5.592318463692739e-06, "loss": 0.0979, "step": 11028 }, { "epoch": 3.42, "learning_rate": 5.591918383676735e-06, "loss": 0.0529, "step": 11029 }, { "epoch": 3.42, "learning_rate": 5.591518303660732e-06, "loss": 0.0374, "step": 11030 }, { "epoch": 3.42, "learning_rate": 5.5911182236447296e-06, "loss": 0.0246, "step": 11031 }, { "epoch": 3.42, "learning_rate": 5.590718143628726e-06, "loss": 0.0356, "step": 11032 }, { "epoch": 3.42, "learning_rate": 5.590318063612723e-06, "loss": 0.0344, "step": 11033 }, { "epoch": 3.42, "learning_rate": 5.58991798359672e-06, "loss": 0.0305, "step": 11034 }, { "epoch": 3.43, "learning_rate": 5.589517903580717e-06, "loss": 0.0289, "step": 11035 }, { "epoch": 3.43, "learning_rate": 5.589117823564714e-06, "loss": 0.0128, "step": 11036 }, { "epoch": 3.43, "learning_rate": 5.5887177435487105e-06, "loss": 0.0284, "step": 11037 }, { "epoch": 3.43, "learning_rate": 5.588317663532707e-06, "loss": 0.033, "step": 11038 }, { "epoch": 3.43, "learning_rate": 5.587917583516704e-06, "loss": 0.0394, "step": 11039 }, { "epoch": 3.43, "learning_rate": 5.5875175035007e-06, "loss": 0.0413, "step": 11040 }, { "epoch": 3.43, "learning_rate": 5.587117423484697e-06, "loss": 0.0291, "step": 11041 }, { "epoch": 3.43, "learning_rate": 5.586717343468694e-06, "loss": 0.0211, "step": 11042 }, { "epoch": 3.43, "learning_rate": 5.586317263452691e-06, "loss": 0.0295, "step": 11043 }, { "epoch": 3.43, "learning_rate": 5.585917183436687e-06, "loss": 0.0229, "step": 11044 }, { "epoch": 3.43, "learning_rate": 5.585517103420685e-06, "loss": 0.0568, "step": 11045 }, { "epoch": 3.43, "learning_rate": 5.5851170234046816e-06, "loss": 0.0194, "step": 11046 }, { "epoch": 3.43, "learning_rate": 5.584716943388678e-06, "loss": 0.0448, "step": 11047 }, { "epoch": 3.43, "learning_rate": 5.584316863372675e-06, "loss": 0.0345, "step": 11048 }, { "epoch": 3.43, "learning_rate": 5.5839167833566725e-06, "loss": 0.042, "step": 11049 }, { "epoch": 3.43, "learning_rate": 5.583516703340669e-06, "loss": 0.1583, "step": 11050 }, { "epoch": 3.43, "learning_rate": 5.583116623324666e-06, "loss": 0.0251, "step": 11051 }, { "epoch": 3.43, "learning_rate": 5.582716543308662e-06, "loss": 0.0309, "step": 11052 }, { "epoch": 3.43, "learning_rate": 5.582316463292658e-06, "loss": 0.0178, "step": 11053 }, { "epoch": 3.43, "learning_rate": 5.581916383276655e-06, "loss": 0.1093, "step": 11054 }, { "epoch": 3.43, "learning_rate": 5.581516303260653e-06, "loss": 0.0371, "step": 11055 }, { "epoch": 3.43, "learning_rate": 5.581116223244649e-06, "loss": 0.0351, "step": 11056 }, { "epoch": 3.43, "learning_rate": 5.580716143228646e-06, "loss": 0.0653, "step": 11057 }, { "epoch": 3.43, "learning_rate": 5.580316063212643e-06, "loss": 0.0408, "step": 11058 }, { "epoch": 3.43, "learning_rate": 5.57991598319664e-06, "loss": 0.0164, "step": 11059 }, { "epoch": 3.43, "learning_rate": 5.579515903180637e-06, "loss": 0.0204, "step": 11060 }, { "epoch": 3.43, "learning_rate": 5.5791158231646335e-06, "loss": 0.0155, "step": 11061 }, { "epoch": 3.43, "learning_rate": 5.57871574314863e-06, "loss": 0.041, "step": 11062 }, { "epoch": 3.43, "learning_rate": 5.578315663132628e-06, "loss": 0.0376, "step": 11063 }, { "epoch": 3.43, "learning_rate": 5.5779155831166245e-06, "loss": 0.0316, "step": 11064 }, { "epoch": 3.43, "learning_rate": 5.57751550310062e-06, "loss": 0.0376, "step": 11065 }, { "epoch": 3.43, "learning_rate": 5.577115423084617e-06, "loss": 0.0449, "step": 11066 }, { "epoch": 3.43, "learning_rate": 5.576715343068614e-06, "loss": 0.0384, "step": 11067 }, { "epoch": 3.44, "learning_rate": 5.57631526305261e-06, "loss": 0.042, "step": 11068 }, { "epoch": 3.44, "learning_rate": 5.575915183036608e-06, "loss": 0.0424, "step": 11069 }, { "epoch": 3.44, "learning_rate": 5.5755151030206046e-06, "loss": 0.0409, "step": 11070 }, { "epoch": 3.44, "learning_rate": 5.575115023004601e-06, "loss": 0.0318, "step": 11071 }, { "epoch": 3.44, "learning_rate": 5.574714942988598e-06, "loss": 0.0443, "step": 11072 }, { "epoch": 3.44, "learning_rate": 5.5743148629725955e-06, "loss": 0.0374, "step": 11073 }, { "epoch": 3.44, "learning_rate": 5.573914782956592e-06, "loss": 0.0429, "step": 11074 }, { "epoch": 3.44, "learning_rate": 5.573514702940589e-06, "loss": 0.078, "step": 11075 }, { "epoch": 3.44, "learning_rate": 5.5731146229245855e-06, "loss": 0.0747, "step": 11076 }, { "epoch": 3.44, "learning_rate": 5.572714542908583e-06, "loss": 0.0607, "step": 11077 }, { "epoch": 3.44, "learning_rate": 5.572314462892579e-06, "loss": 0.0573, "step": 11078 }, { "epoch": 3.44, "learning_rate": 5.571914382876576e-06, "loss": 0.0309, "step": 11079 }, { "epoch": 3.44, "learning_rate": 5.571514302860572e-06, "loss": 0.0718, "step": 11080 }, { "epoch": 3.44, "learning_rate": 5.571114222844569e-06, "loss": 0.0577, "step": 11081 }, { "epoch": 3.44, "learning_rate": 5.570714142828566e-06, "loss": 0.0532, "step": 11082 }, { "epoch": 3.44, "learning_rate": 5.570314062812563e-06, "loss": 0.0563, "step": 11083 }, { "epoch": 3.44, "learning_rate": 5.56991398279656e-06, "loss": 0.0125, "step": 11084 }, { "epoch": 3.44, "learning_rate": 5.5695139027805566e-06, "loss": 0.0428, "step": 11085 }, { "epoch": 3.44, "learning_rate": 5.569113822764553e-06, "loss": 0.0807, "step": 11086 }, { "epoch": 3.44, "learning_rate": 5.568713742748551e-06, "loss": 0.0367, "step": 11087 }, { "epoch": 3.44, "learning_rate": 5.5683136627325475e-06, "loss": 0.0394, "step": 11088 }, { "epoch": 3.44, "learning_rate": 5.567913582716544e-06, "loss": 0.0621, "step": 11089 }, { "epoch": 3.44, "learning_rate": 5.56751350270054e-06, "loss": 0.068, "step": 11090 }, { "epoch": 3.44, "learning_rate": 5.567113422684537e-06, "loss": 0.0304, "step": 11091 }, { "epoch": 3.44, "learning_rate": 5.566713342668534e-06, "loss": 0.046, "step": 11092 }, { "epoch": 3.44, "learning_rate": 5.566313262652531e-06, "loss": 0.0839, "step": 11093 }, { "epoch": 3.44, "learning_rate": 5.565913182636528e-06, "loss": 0.0342, "step": 11094 }, { "epoch": 3.44, "learning_rate": 5.565513102620524e-06, "loss": 0.0373, "step": 11095 }, { "epoch": 3.44, "learning_rate": 5.565113022604521e-06, "loss": 0.0108, "step": 11096 }, { "epoch": 3.44, "learning_rate": 5.5647129425885185e-06, "loss": 0.1351, "step": 11097 }, { "epoch": 3.44, "learning_rate": 5.564312862572515e-06, "loss": 0.0261, "step": 11098 }, { "epoch": 3.44, "learning_rate": 5.563912782556512e-06, "loss": 0.1095, "step": 11099 }, { "epoch": 3.45, "learning_rate": 5.5635127025405085e-06, "loss": 0.0169, "step": 11100 }, { "epoch": 3.45, "learning_rate": 5.563112622524506e-06, "loss": 0.0412, "step": 11101 }, { "epoch": 3.45, "learning_rate": 5.562712542508503e-06, "loss": 0.1453, "step": 11102 }, { "epoch": 3.45, "learning_rate": 5.562312462492499e-06, "loss": 0.0672, "step": 11103 }, { "epoch": 3.45, "learning_rate": 5.561912382476495e-06, "loss": 0.0337, "step": 11104 }, { "epoch": 3.45, "learning_rate": 5.561512302460492e-06, "loss": 0.0734, "step": 11105 }, { "epoch": 3.45, "learning_rate": 5.5611122224444895e-06, "loss": 0.0399, "step": 11106 }, { "epoch": 3.45, "learning_rate": 5.560712142428486e-06, "loss": 0.0228, "step": 11107 }, { "epoch": 3.45, "learning_rate": 5.560312062412483e-06, "loss": 0.0354, "step": 11108 }, { "epoch": 3.45, "learning_rate": 5.5599119823964796e-06, "loss": 0.043, "step": 11109 }, { "epoch": 3.45, "learning_rate": 5.559511902380476e-06, "loss": 0.0263, "step": 11110 }, { "epoch": 3.45, "learning_rate": 5.559111822364474e-06, "loss": 0.0291, "step": 11111 }, { "epoch": 3.45, "learning_rate": 5.5587117423484705e-06, "loss": 0.024, "step": 11112 }, { "epoch": 3.45, "learning_rate": 5.558311662332467e-06, "loss": 0.0363, "step": 11113 }, { "epoch": 3.45, "learning_rate": 5.557911582316464e-06, "loss": 0.0375, "step": 11114 }, { "epoch": 3.45, "learning_rate": 5.55751150230046e-06, "loss": 0.0728, "step": 11115 }, { "epoch": 3.45, "learning_rate": 5.557111422284457e-06, "loss": 0.0485, "step": 11116 }, { "epoch": 3.45, "learning_rate": 5.556711342268454e-06, "loss": 0.0447, "step": 11117 }, { "epoch": 3.45, "learning_rate": 5.556311262252451e-06, "loss": 0.0151, "step": 11118 }, { "epoch": 3.45, "learning_rate": 5.555911182236447e-06, "loss": 0.0435, "step": 11119 }, { "epoch": 3.45, "learning_rate": 5.555511102220445e-06, "loss": 0.0325, "step": 11120 }, { "epoch": 3.45, "learning_rate": 5.5551110222044415e-06, "loss": 0.0374, "step": 11121 }, { "epoch": 3.45, "learning_rate": 5.554710942188438e-06, "loss": 0.0437, "step": 11122 }, { "epoch": 3.45, "learning_rate": 5.554310862172435e-06, "loss": 0.0244, "step": 11123 }, { "epoch": 3.45, "learning_rate": 5.5539107821564316e-06, "loss": 0.0354, "step": 11124 }, { "epoch": 3.45, "learning_rate": 5.553510702140429e-06, "loss": 0.0148, "step": 11125 }, { "epoch": 3.45, "learning_rate": 5.553110622124426e-06, "loss": 0.1186, "step": 11126 }, { "epoch": 3.45, "learning_rate": 5.5527105421084225e-06, "loss": 0.0296, "step": 11127 }, { "epoch": 3.45, "learning_rate": 5.552310462092418e-06, "loss": 0.085, "step": 11128 }, { "epoch": 3.45, "learning_rate": 5.551910382076415e-06, "loss": 0.0303, "step": 11129 }, { "epoch": 3.45, "learning_rate": 5.5515103020604125e-06, "loss": 0.0408, "step": 11130 }, { "epoch": 3.45, "learning_rate": 5.551110222044409e-06, "loss": 0.0552, "step": 11131 }, { "epoch": 3.46, "learning_rate": 5.550710142028406e-06, "loss": 0.0182, "step": 11132 }, { "epoch": 3.46, "learning_rate": 5.550310062012403e-06, "loss": 0.039, "step": 11133 }, { "epoch": 3.46, "learning_rate": 5.5499099819964e-06, "loss": 0.0311, "step": 11134 }, { "epoch": 3.46, "learning_rate": 5.549509901980397e-06, "loss": 0.0425, "step": 11135 }, { "epoch": 3.46, "learning_rate": 5.5491098219643935e-06, "loss": 0.0736, "step": 11136 }, { "epoch": 3.46, "learning_rate": 5.54870974194839e-06, "loss": 0.0488, "step": 11137 }, { "epoch": 3.46, "learning_rate": 5.548309661932388e-06, "loss": 0.0338, "step": 11138 }, { "epoch": 3.46, "learning_rate": 5.547909581916384e-06, "loss": 0.0289, "step": 11139 }, { "epoch": 3.46, "learning_rate": 5.547509501900381e-06, "loss": 0.0438, "step": 11140 }, { "epoch": 3.46, "learning_rate": 5.547109421884377e-06, "loss": 0.0772, "step": 11141 }, { "epoch": 3.46, "learning_rate": 5.546709341868374e-06, "loss": 0.0633, "step": 11142 }, { "epoch": 3.46, "learning_rate": 5.54630926185237e-06, "loss": 0.0311, "step": 11143 }, { "epoch": 3.46, "learning_rate": 5.545909181836368e-06, "loss": 0.0273, "step": 11144 }, { "epoch": 3.46, "learning_rate": 5.5455091018203645e-06, "loss": 0.0742, "step": 11145 }, { "epoch": 3.46, "learning_rate": 5.545109021804361e-06, "loss": 0.046, "step": 11146 }, { "epoch": 3.46, "learning_rate": 5.544708941788358e-06, "loss": 0.0178, "step": 11147 }, { "epoch": 3.46, "learning_rate": 5.544308861772355e-06, "loss": 0.0254, "step": 11148 }, { "epoch": 3.46, "learning_rate": 5.543908781756352e-06, "loss": 0.0824, "step": 11149 }, { "epoch": 3.46, "learning_rate": 5.543508701740349e-06, "loss": 0.0595, "step": 11150 }, { "epoch": 3.46, "learning_rate": 5.5431086217243455e-06, "loss": 0.0218, "step": 11151 }, { "epoch": 3.46, "learning_rate": 5.542708541708343e-06, "loss": 0.0297, "step": 11152 }, { "epoch": 3.46, "learning_rate": 5.542308461692338e-06, "loss": 0.03, "step": 11153 }, { "epoch": 3.46, "learning_rate": 5.5419083816763355e-06, "loss": 0.0531, "step": 11154 }, { "epoch": 3.46, "learning_rate": 5.541508301660332e-06, "loss": 0.0174, "step": 11155 }, { "epoch": 3.46, "learning_rate": 5.541108221644329e-06, "loss": 0.0171, "step": 11156 }, { "epoch": 3.46, "learning_rate": 5.540708141628326e-06, "loss": 0.0271, "step": 11157 }, { "epoch": 3.46, "learning_rate": 5.540308061612323e-06, "loss": 0.0409, "step": 11158 }, { "epoch": 3.46, "learning_rate": 5.53990798159632e-06, "loss": 0.1353, "step": 11159 }, { "epoch": 3.46, "learning_rate": 5.5395079015803165e-06, "loss": 0.0819, "step": 11160 }, { "epoch": 3.46, "learning_rate": 5.539107821564313e-06, "loss": 0.0781, "step": 11161 }, { "epoch": 3.46, "learning_rate": 5.538707741548311e-06, "loss": 0.05, "step": 11162 }, { "epoch": 3.46, "learning_rate": 5.538307661532307e-06, "loss": 0.0253, "step": 11163 }, { "epoch": 3.47, "learning_rate": 5.537907581516304e-06, "loss": 0.0417, "step": 11164 }, { "epoch": 3.47, "learning_rate": 5.537507501500301e-06, "loss": 0.0327, "step": 11165 }, { "epoch": 3.47, "learning_rate": 5.537107421484297e-06, "loss": 0.0395, "step": 11166 }, { "epoch": 3.47, "learning_rate": 5.536707341468293e-06, "loss": 0.0171, "step": 11167 }, { "epoch": 3.47, "learning_rate": 5.536307261452291e-06, "loss": 0.0383, "step": 11168 }, { "epoch": 3.47, "learning_rate": 5.5359071814362875e-06, "loss": 0.0259, "step": 11169 }, { "epoch": 3.47, "learning_rate": 5.535507101420284e-06, "loss": 0.1148, "step": 11170 }, { "epoch": 3.47, "learning_rate": 5.535107021404281e-06, "loss": 0.0201, "step": 11171 }, { "epoch": 3.47, "learning_rate": 5.5347069413882784e-06, "loss": 0.0413, "step": 11172 }, { "epoch": 3.47, "learning_rate": 5.534306861372275e-06, "loss": 0.0385, "step": 11173 }, { "epoch": 3.47, "learning_rate": 5.533906781356272e-06, "loss": 0.0556, "step": 11174 }, { "epoch": 3.47, "learning_rate": 5.5335067013402685e-06, "loss": 0.0373, "step": 11175 }, { "epoch": 3.47, "learning_rate": 5.533106621324266e-06, "loss": 0.0442, "step": 11176 }, { "epoch": 3.47, "learning_rate": 5.532706541308263e-06, "loss": 0.047, "step": 11177 }, { "epoch": 3.47, "learning_rate": 5.532306461292259e-06, "loss": 0.0342, "step": 11178 }, { "epoch": 3.47, "learning_rate": 5.531906381276255e-06, "loss": 0.116, "step": 11179 }, { "epoch": 3.47, "learning_rate": 5.531506301260252e-06, "loss": 0.1542, "step": 11180 }, { "epoch": 3.47, "learning_rate": 5.531106221244249e-06, "loss": 0.014, "step": 11181 }, { "epoch": 3.47, "learning_rate": 5.530706141228246e-06, "loss": 0.0924, "step": 11182 }, { "epoch": 3.47, "learning_rate": 5.530306061212243e-06, "loss": 0.0269, "step": 11183 }, { "epoch": 3.47, "learning_rate": 5.5299059811962395e-06, "loss": 0.0748, "step": 11184 }, { "epoch": 3.47, "learning_rate": 5.529505901180236e-06, "loss": 0.0144, "step": 11185 }, { "epoch": 3.47, "learning_rate": 5.529105821164234e-06, "loss": 0.0257, "step": 11186 }, { "epoch": 3.47, "learning_rate": 5.52870574114823e-06, "loss": 0.0475, "step": 11187 }, { "epoch": 3.47, "learning_rate": 5.528305661132227e-06, "loss": 0.0187, "step": 11188 }, { "epoch": 3.47, "learning_rate": 5.527905581116224e-06, "loss": 0.0415, "step": 11189 }, { "epoch": 3.47, "learning_rate": 5.527505501100221e-06, "loss": 0.0233, "step": 11190 }, { "epoch": 3.47, "learning_rate": 5.527105421084217e-06, "loss": 0.0506, "step": 11191 }, { "epoch": 3.47, "learning_rate": 5.526705341068214e-06, "loss": 0.0802, "step": 11192 }, { "epoch": 3.47, "learning_rate": 5.5263052610522105e-06, "loss": 0.0586, "step": 11193 }, { "epoch": 3.47, "learning_rate": 5.525905181036207e-06, "loss": 0.0577, "step": 11194 }, { "epoch": 3.47, "learning_rate": 5.525505101020204e-06, "loss": 0.0488, "step": 11195 }, { "epoch": 3.47, "learning_rate": 5.5251050210042014e-06, "loss": 0.0254, "step": 11196 }, { "epoch": 3.48, "learning_rate": 5.524704940988198e-06, "loss": 0.0582, "step": 11197 }, { "epoch": 3.48, "learning_rate": 5.524304860972195e-06, "loss": 0.1831, "step": 11198 }, { "epoch": 3.48, "learning_rate": 5.5239047809561915e-06, "loss": 0.0609, "step": 11199 }, { "epoch": 3.48, "learning_rate": 5.523504700940189e-06, "loss": 0.0652, "step": 11200 }, { "epoch": 3.48, "learning_rate": 5.523104620924186e-06, "loss": 0.0365, "step": 11201 }, { "epoch": 3.48, "learning_rate": 5.522704540908182e-06, "loss": 0.0394, "step": 11202 }, { "epoch": 3.48, "learning_rate": 5.522304460892179e-06, "loss": 0.0603, "step": 11203 }, { "epoch": 3.48, "learning_rate": 5.521904380876175e-06, "loss": 0.0297, "step": 11204 }, { "epoch": 3.48, "learning_rate": 5.5215043008601725e-06, "loss": 0.0319, "step": 11205 }, { "epoch": 3.48, "learning_rate": 5.521104220844169e-06, "loss": 0.037, "step": 11206 }, { "epoch": 3.48, "learning_rate": 5.520704140828166e-06, "loss": 0.0278, "step": 11207 }, { "epoch": 3.48, "learning_rate": 5.5203040608121625e-06, "loss": 0.0471, "step": 11208 }, { "epoch": 3.48, "learning_rate": 5.51990398079616e-06, "loss": 0.0236, "step": 11209 }, { "epoch": 3.48, "learning_rate": 5.519503900780157e-06, "loss": 0.0411, "step": 11210 }, { "epoch": 3.48, "learning_rate": 5.5191038207641534e-06, "loss": 0.0417, "step": 11211 }, { "epoch": 3.48, "learning_rate": 5.51870374074815e-06, "loss": 0.0697, "step": 11212 }, { "epoch": 3.48, "learning_rate": 5.518303660732147e-06, "loss": 0.0302, "step": 11213 }, { "epoch": 3.48, "learning_rate": 5.517903580716144e-06, "loss": 0.1185, "step": 11214 }, { "epoch": 3.48, "learning_rate": 5.517503500700141e-06, "loss": 0.0468, "step": 11215 }, { "epoch": 3.48, "learning_rate": 5.517103420684137e-06, "loss": 0.0349, "step": 11216 }, { "epoch": 3.48, "learning_rate": 5.5167033406681335e-06, "loss": 0.08, "step": 11217 }, { "epoch": 3.48, "learning_rate": 5.51630326065213e-06, "loss": 0.0441, "step": 11218 }, { "epoch": 3.48, "learning_rate": 5.515903180636128e-06, "loss": 0.0267, "step": 11219 }, { "epoch": 3.48, "learning_rate": 5.5155031006201245e-06, "loss": 0.1091, "step": 11220 }, { "epoch": 3.48, "learning_rate": 5.515103020604121e-06, "loss": 0.035, "step": 11221 }, { "epoch": 3.48, "learning_rate": 5.514702940588118e-06, "loss": 0.0514, "step": 11222 }, { "epoch": 3.48, "learning_rate": 5.514302860572115e-06, "loss": 0.0264, "step": 11223 }, { "epoch": 3.48, "learning_rate": 5.513902780556112e-06, "loss": 0.0661, "step": 11224 }, { "epoch": 3.48, "learning_rate": 5.513502700540109e-06, "loss": 0.0467, "step": 11225 }, { "epoch": 3.48, "learning_rate": 5.513102620524105e-06, "loss": 0.0544, "step": 11226 }, { "epoch": 3.48, "learning_rate": 5.512702540508102e-06, "loss": 0.0439, "step": 11227 }, { "epoch": 3.48, "learning_rate": 5.5123024604921e-06, "loss": 0.1408, "step": 11228 }, { "epoch": 3.49, "learning_rate": 5.5119023804760955e-06, "loss": 0.0237, "step": 11229 }, { "epoch": 3.49, "learning_rate": 5.511502300460092e-06, "loss": 0.0459, "step": 11230 }, { "epoch": 3.49, "learning_rate": 5.511102220444089e-06, "loss": 0.1565, "step": 11231 }, { "epoch": 3.49, "learning_rate": 5.5107021404280855e-06, "loss": 0.0479, "step": 11232 }, { "epoch": 3.49, "learning_rate": 5.510302060412083e-06, "loss": 0.0329, "step": 11233 }, { "epoch": 3.49, "learning_rate": 5.50990198039608e-06, "loss": 0.0313, "step": 11234 }, { "epoch": 3.49, "learning_rate": 5.5095019003800764e-06, "loss": 0.0146, "step": 11235 }, { "epoch": 3.49, "learning_rate": 5.509101820364073e-06, "loss": 0.0307, "step": 11236 }, { "epoch": 3.49, "learning_rate": 5.508701740348071e-06, "loss": 0.0495, "step": 11237 }, { "epoch": 3.49, "learning_rate": 5.508301660332067e-06, "loss": 0.0173, "step": 11238 }, { "epoch": 3.49, "learning_rate": 5.507901580316064e-06, "loss": 0.0352, "step": 11239 }, { "epoch": 3.49, "learning_rate": 5.507501500300061e-06, "loss": 0.0229, "step": 11240 }, { "epoch": 3.49, "learning_rate": 5.507101420284057e-06, "loss": 0.0215, "step": 11241 }, { "epoch": 3.49, "learning_rate": 5.506701340268053e-06, "loss": 0.0263, "step": 11242 }, { "epoch": 3.49, "learning_rate": 5.506301260252051e-06, "loss": 0.0306, "step": 11243 }, { "epoch": 3.49, "learning_rate": 5.5059011802360475e-06, "loss": 0.0184, "step": 11244 }, { "epoch": 3.49, "learning_rate": 5.505501100220044e-06, "loss": 0.0416, "step": 11245 }, { "epoch": 3.49, "learning_rate": 5.505101020204041e-06, "loss": 0.0483, "step": 11246 }, { "epoch": 3.49, "learning_rate": 5.504700940188038e-06, "loss": 0.082, "step": 11247 }, { "epoch": 3.49, "learning_rate": 5.504300860172035e-06, "loss": 0.0222, "step": 11248 }, { "epoch": 3.49, "learning_rate": 5.503900780156032e-06, "loss": 0.0611, "step": 11249 }, { "epoch": 3.49, "learning_rate": 5.5035007001400284e-06, "loss": 0.0977, "step": 11250 }, { "epoch": 3.49, "learning_rate": 5.503100620124026e-06, "loss": 0.0296, "step": 11251 }, { "epoch": 3.49, "learning_rate": 5.502700540108023e-06, "loss": 0.0393, "step": 11252 }, { "epoch": 3.49, "learning_rate": 5.502300460092019e-06, "loss": 0.0652, "step": 11253 }, { "epoch": 3.49, "learning_rate": 5.501900380076015e-06, "loss": 0.0578, "step": 11254 }, { "epoch": 3.49, "learning_rate": 5.501500300060012e-06, "loss": 0.0788, "step": 11255 }, { "epoch": 3.49, "learning_rate": 5.5011002200440085e-06, "loss": 0.0261, "step": 11256 }, { "epoch": 3.49, "learning_rate": 5.500700140028006e-06, "loss": 0.0202, "step": 11257 }, { "epoch": 3.49, "learning_rate": 5.500300060012003e-06, "loss": 0.0542, "step": 11258 }, { "epoch": 3.49, "learning_rate": 5.4998999799959995e-06, "loss": 0.0436, "step": 11259 }, { "epoch": 3.49, "learning_rate": 5.499499899979996e-06, "loss": 0.069, "step": 11260 }, { "epoch": 3.5, "learning_rate": 5.499099819963994e-06, "loss": 0.0291, "step": 11261 }, { "epoch": 3.5, "learning_rate": 5.49869973994799e-06, "loss": 0.0383, "step": 11262 }, { "epoch": 3.5, "learning_rate": 5.498299659931987e-06, "loss": 0.0575, "step": 11263 }, { "epoch": 3.5, "learning_rate": 5.497899579915984e-06, "loss": 0.0414, "step": 11264 }, { "epoch": 3.5, "learning_rate": 5.497499499899981e-06, "loss": 0.2096, "step": 11265 }, { "epoch": 3.5, "learning_rate": 5.497099419883978e-06, "loss": 0.0407, "step": 11266 }, { "epoch": 3.5, "learning_rate": 5.496699339867974e-06, "loss": 0.0301, "step": 11267 }, { "epoch": 3.5, "learning_rate": 5.4962992598519705e-06, "loss": 0.0394, "step": 11268 }, { "epoch": 3.5, "learning_rate": 5.495899179835967e-06, "loss": 0.048, "step": 11269 }, { "epoch": 3.5, "learning_rate": 5.495499099819964e-06, "loss": 0.0194, "step": 11270 }, { "epoch": 3.5, "learning_rate": 5.495099019803961e-06, "loss": 0.0615, "step": 11271 }, { "epoch": 3.5, "learning_rate": 5.494698939787958e-06, "loss": 0.0278, "step": 11272 }, { "epoch": 3.5, "learning_rate": 5.494298859771955e-06, "loss": 0.0261, "step": 11273 }, { "epoch": 3.5, "learning_rate": 5.4938987797559514e-06, "loss": 0.0512, "step": 11274 }, { "epoch": 3.5, "learning_rate": 5.493498699739949e-06, "loss": 0.0897, "step": 11275 }, { "epoch": 3.5, "learning_rate": 5.493098619723946e-06, "loss": 0.0282, "step": 11276 }, { "epoch": 3.5, "learning_rate": 5.492698539707942e-06, "loss": 0.0232, "step": 11277 }, { "epoch": 3.5, "learning_rate": 5.492298459691939e-06, "loss": 0.0787, "step": 11278 }, { "epoch": 3.5, "learning_rate": 5.491898379675935e-06, "loss": 0.0271, "step": 11279 }, { "epoch": 3.5, "learning_rate": 5.4914982996599316e-06, "loss": 0.0546, "step": 11280 }, { "epoch": 3.5, "learning_rate": 5.491098219643929e-06, "loss": 0.0297, "step": 11281 }, { "epoch": 3.5, "learning_rate": 5.490698139627926e-06, "loss": 0.0178, "step": 11282 }, { "epoch": 3.5, "learning_rate": 5.4902980596119225e-06, "loss": 0.0237, "step": 11283 }, { "epoch": 3.5, "learning_rate": 5.489897979595919e-06, "loss": 0.1595, "step": 11284 }, { "epoch": 3.5, "learning_rate": 5.489497899579917e-06, "loss": 0.0211, "step": 11285 }, { "epoch": 3.5, "learning_rate": 5.489097819563913e-06, "loss": 0.1107, "step": 11286 }, { "epoch": 3.5, "learning_rate": 5.48869773954791e-06, "loss": 0.0477, "step": 11287 }, { "epoch": 3.5, "learning_rate": 5.488297659531907e-06, "loss": 0.0329, "step": 11288 }, { "epoch": 3.5, "learning_rate": 5.487897579515904e-06, "loss": 0.0436, "step": 11289 }, { "epoch": 3.5, "learning_rate": 5.487497499499901e-06, "loss": 0.0953, "step": 11290 }, { "epoch": 3.5, "learning_rate": 5.487097419483898e-06, "loss": 0.105, "step": 11291 }, { "epoch": 3.5, "learning_rate": 5.4866973394678935e-06, "loss": 0.0477, "step": 11292 }, { "epoch": 3.51, "learning_rate": 5.48629725945189e-06, "loss": 0.06, "step": 11293 }, { "epoch": 3.51, "learning_rate": 5.485897179435888e-06, "loss": 0.021, "step": 11294 }, { "epoch": 3.51, "learning_rate": 5.485497099419884e-06, "loss": 0.0355, "step": 11295 }, { "epoch": 3.51, "learning_rate": 5.485097019403881e-06, "loss": 0.0603, "step": 11296 }, { "epoch": 3.51, "learning_rate": 5.484696939387878e-06, "loss": 0.04, "step": 11297 }, { "epoch": 3.51, "learning_rate": 5.4842968593718745e-06, "loss": 0.0489, "step": 11298 }, { "epoch": 3.51, "learning_rate": 5.483896779355872e-06, "loss": 0.0435, "step": 11299 }, { "epoch": 3.51, "learning_rate": 5.483496699339869e-06, "loss": 0.0533, "step": 11300 }, { "epoch": 3.51, "learning_rate": 5.483096619323865e-06, "loss": 0.0531, "step": 11301 }, { "epoch": 3.51, "learning_rate": 5.482696539307862e-06, "loss": 0.0287, "step": 11302 }, { "epoch": 3.51, "learning_rate": 5.48229645929186e-06, "loss": 0.0335, "step": 11303 }, { "epoch": 3.51, "learning_rate": 5.481896379275856e-06, "loss": 0.0632, "step": 11304 }, { "epoch": 3.51, "learning_rate": 5.481496299259852e-06, "loss": 0.0511, "step": 11305 }, { "epoch": 3.51, "learning_rate": 5.481096219243849e-06, "loss": 0.0202, "step": 11306 }, { "epoch": 3.51, "learning_rate": 5.4806961392278455e-06, "loss": 0.0429, "step": 11307 }, { "epoch": 3.51, "learning_rate": 5.480296059211843e-06, "loss": 0.0655, "step": 11308 }, { "epoch": 3.51, "learning_rate": 5.47989597919584e-06, "loss": 0.041, "step": 11309 }, { "epoch": 3.51, "learning_rate": 5.479495899179836e-06, "loss": 0.0998, "step": 11310 }, { "epoch": 3.51, "learning_rate": 5.479095819163833e-06, "loss": 0.1305, "step": 11311 }, { "epoch": 3.51, "learning_rate": 5.47869573914783e-06, "loss": 0.0815, "step": 11312 }, { "epoch": 3.51, "learning_rate": 5.478295659131827e-06, "loss": 0.0559, "step": 11313 }, { "epoch": 3.51, "learning_rate": 5.477895579115824e-06, "loss": 0.0492, "step": 11314 }, { "epoch": 3.51, "learning_rate": 5.477495499099821e-06, "loss": 0.0308, "step": 11315 }, { "epoch": 3.51, "learning_rate": 5.477095419083817e-06, "loss": 0.0194, "step": 11316 }, { "epoch": 3.51, "learning_rate": 5.476695339067813e-06, "loss": 0.0309, "step": 11317 }, { "epoch": 3.51, "learning_rate": 5.476295259051811e-06, "loss": 0.0489, "step": 11318 }, { "epoch": 3.51, "learning_rate": 5.475895179035807e-06, "loss": 0.0292, "step": 11319 }, { "epoch": 3.51, "learning_rate": 5.475495099019804e-06, "loss": 0.0502, "step": 11320 }, { "epoch": 3.51, "learning_rate": 5.475095019003801e-06, "loss": 0.0335, "step": 11321 }, { "epoch": 3.51, "learning_rate": 5.474694938987798e-06, "loss": 0.0237, "step": 11322 }, { "epoch": 3.51, "learning_rate": 5.474294858971795e-06, "loss": 0.0463, "step": 11323 }, { "epoch": 3.51, "learning_rate": 5.473894778955792e-06, "loss": 0.0305, "step": 11324 }, { "epoch": 3.52, "learning_rate": 5.473494698939788e-06, "loss": 0.038, "step": 11325 }, { "epoch": 3.52, "learning_rate": 5.473094618923785e-06, "loss": 0.0493, "step": 11326 }, { "epoch": 3.52, "learning_rate": 5.472694538907783e-06, "loss": 0.0591, "step": 11327 }, { "epoch": 3.52, "learning_rate": 5.472294458891779e-06, "loss": 0.0172, "step": 11328 }, { "epoch": 3.52, "learning_rate": 5.471894378875776e-06, "loss": 0.0376, "step": 11329 }, { "epoch": 3.52, "learning_rate": 5.471494298859772e-06, "loss": 0.0198, "step": 11330 }, { "epoch": 3.52, "learning_rate": 5.4710942188437685e-06, "loss": 0.0469, "step": 11331 }, { "epoch": 3.52, "learning_rate": 5.470694138827766e-06, "loss": 0.0444, "step": 11332 }, { "epoch": 3.52, "learning_rate": 5.470294058811763e-06, "loss": 0.0261, "step": 11333 }, { "epoch": 3.52, "learning_rate": 5.469893978795759e-06, "loss": 0.0356, "step": 11334 }, { "epoch": 3.52, "learning_rate": 5.469493898779756e-06, "loss": 0.0211, "step": 11335 }, { "epoch": 3.52, "learning_rate": 5.469093818763754e-06, "loss": 0.0307, "step": 11336 }, { "epoch": 3.52, "learning_rate": 5.46869373874775e-06, "loss": 0.0683, "step": 11337 }, { "epoch": 3.52, "learning_rate": 5.468293658731747e-06, "loss": 0.0674, "step": 11338 }, { "epoch": 3.52, "learning_rate": 5.467893578715744e-06, "loss": 0.0788, "step": 11339 }, { "epoch": 3.52, "learning_rate": 5.46749349869974e-06, "loss": 0.0495, "step": 11340 }, { "epoch": 3.52, "learning_rate": 5.467093418683738e-06, "loss": 0.021, "step": 11341 }, { "epoch": 3.52, "learning_rate": 5.466693338667734e-06, "loss": 0.0256, "step": 11342 }, { "epoch": 3.52, "learning_rate": 5.46629325865173e-06, "loss": 0.0391, "step": 11343 }, { "epoch": 3.52, "learning_rate": 5.465893178635727e-06, "loss": 0.0701, "step": 11344 }, { "epoch": 3.52, "learning_rate": 5.465493098619724e-06, "loss": 0.1169, "step": 11345 }, { "epoch": 3.52, "learning_rate": 5.465093018603721e-06, "loss": 0.1003, "step": 11346 }, { "epoch": 3.52, "learning_rate": 5.464692938587718e-06, "loss": 0.0414, "step": 11347 }, { "epoch": 3.52, "learning_rate": 5.464292858571715e-06, "loss": 0.1268, "step": 11348 }, { "epoch": 3.52, "learning_rate": 5.463892778555711e-06, "loss": 0.0895, "step": 11349 }, { "epoch": 3.52, "learning_rate": 5.463492698539709e-06, "loss": 0.0627, "step": 11350 }, { "epoch": 3.52, "learning_rate": 5.463092618523706e-06, "loss": 0.0404, "step": 11351 }, { "epoch": 3.52, "learning_rate": 5.462692538507702e-06, "loss": 0.0345, "step": 11352 }, { "epoch": 3.52, "learning_rate": 5.462292458491699e-06, "loss": 0.0263, "step": 11353 }, { "epoch": 3.52, "learning_rate": 5.4618923784756965e-06, "loss": 0.0283, "step": 11354 }, { "epoch": 3.52, "learning_rate": 5.4614922984596915e-06, "loss": 0.0541, "step": 11355 }, { "epoch": 3.52, "learning_rate": 5.461092218443689e-06, "loss": 0.031, "step": 11356 }, { "epoch": 3.52, "learning_rate": 5.460692138427686e-06, "loss": 0.0514, "step": 11357 }, { "epoch": 3.53, "learning_rate": 5.460292058411682e-06, "loss": 0.0148, "step": 11358 }, { "epoch": 3.53, "learning_rate": 5.459891978395679e-06, "loss": 0.0683, "step": 11359 }, { "epoch": 3.53, "learning_rate": 5.459491898379677e-06, "loss": 0.0322, "step": 11360 }, { "epoch": 3.53, "learning_rate": 5.459091818363673e-06, "loss": 0.0357, "step": 11361 }, { "epoch": 3.53, "learning_rate": 5.45869173834767e-06, "loss": 0.048, "step": 11362 }, { "epoch": 3.53, "learning_rate": 5.458291658331667e-06, "loss": 0.0967, "step": 11363 }, { "epoch": 3.53, "learning_rate": 5.457891578315664e-06, "loss": 0.0339, "step": 11364 }, { "epoch": 3.53, "learning_rate": 5.457491498299661e-06, "loss": 0.0284, "step": 11365 }, { "epoch": 3.53, "learning_rate": 5.457091418283658e-06, "loss": 0.0187, "step": 11366 }, { "epoch": 3.53, "learning_rate": 5.456691338267654e-06, "loss": 0.0311, "step": 11367 }, { "epoch": 3.53, "learning_rate": 5.45629125825165e-06, "loss": 0.2037, "step": 11368 }, { "epoch": 3.53, "learning_rate": 5.455891178235647e-06, "loss": 0.0552, "step": 11369 }, { "epoch": 3.53, "learning_rate": 5.455491098219644e-06, "loss": 0.0365, "step": 11370 }, { "epoch": 3.53, "learning_rate": 5.455091018203641e-06, "loss": 0.0372, "step": 11371 }, { "epoch": 3.53, "learning_rate": 5.454690938187638e-06, "loss": 0.0439, "step": 11372 }, { "epoch": 3.53, "learning_rate": 5.454290858171634e-06, "loss": 0.0363, "step": 11373 }, { "epoch": 3.53, "learning_rate": 5.453890778155632e-06, "loss": 0.0513, "step": 11374 }, { "epoch": 3.53, "learning_rate": 5.453490698139629e-06, "loss": 0.0234, "step": 11375 }, { "epoch": 3.53, "learning_rate": 5.453090618123625e-06, "loss": 0.0613, "step": 11376 }, { "epoch": 3.53, "learning_rate": 5.452690538107622e-06, "loss": 0.0177, "step": 11377 }, { "epoch": 3.53, "learning_rate": 5.4522904580916195e-06, "loss": 0.0296, "step": 11378 }, { "epoch": 3.53, "learning_rate": 5.451890378075616e-06, "loss": 0.0403, "step": 11379 }, { "epoch": 3.53, "learning_rate": 5.451490298059612e-06, "loss": 0.0659, "step": 11380 }, { "epoch": 3.53, "learning_rate": 5.451090218043609e-06, "loss": 0.0324, "step": 11381 }, { "epoch": 3.53, "learning_rate": 5.450690138027605e-06, "loss": 0.0435, "step": 11382 }, { "epoch": 3.53, "learning_rate": 5.450290058011602e-06, "loss": 0.0374, "step": 11383 }, { "epoch": 3.53, "learning_rate": 5.4498899779956e-06, "loss": 0.111, "step": 11384 }, { "epoch": 3.53, "learning_rate": 5.449489897979596e-06, "loss": 0.031, "step": 11385 }, { "epoch": 3.53, "learning_rate": 5.449089817963593e-06, "loss": 0.0314, "step": 11386 }, { "epoch": 3.53, "learning_rate": 5.44868973794759e-06, "loss": 0.1501, "step": 11387 }, { "epoch": 3.53, "learning_rate": 5.448289657931587e-06, "loss": 0.0334, "step": 11388 }, { "epoch": 3.53, "learning_rate": 5.447889577915584e-06, "loss": 0.0785, "step": 11389 }, { "epoch": 3.54, "learning_rate": 5.447489497899581e-06, "loss": 0.0159, "step": 11390 }, { "epoch": 3.54, "learning_rate": 5.447089417883577e-06, "loss": 0.0415, "step": 11391 }, { "epoch": 3.54, "learning_rate": 5.446689337867575e-06, "loss": 0.018, "step": 11392 }, { "epoch": 3.54, "learning_rate": 5.446289257851571e-06, "loss": 0.0539, "step": 11393 }, { "epoch": 3.54, "learning_rate": 5.445889177835567e-06, "loss": 0.0129, "step": 11394 }, { "epoch": 3.54, "learning_rate": 5.445489097819564e-06, "loss": 0.056, "step": 11395 }, { "epoch": 3.54, "learning_rate": 5.445089017803561e-06, "loss": 0.0908, "step": 11396 }, { "epoch": 3.54, "learning_rate": 5.444688937787557e-06, "loss": 0.0379, "step": 11397 }, { "epoch": 3.54, "learning_rate": 5.444288857771555e-06, "loss": 0.0333, "step": 11398 }, { "epoch": 3.54, "learning_rate": 5.443888777755552e-06, "loss": 0.0747, "step": 11399 }, { "epoch": 3.54, "learning_rate": 5.443488697739548e-06, "loss": 0.0147, "step": 11400 }, { "epoch": 3.54, "learning_rate": 5.443088617723545e-06, "loss": 0.0509, "step": 11401 }, { "epoch": 3.54, "learning_rate": 5.4426885377075425e-06, "loss": 0.0194, "step": 11402 }, { "epoch": 3.54, "learning_rate": 5.442288457691539e-06, "loss": 0.0654, "step": 11403 }, { "epoch": 3.54, "learning_rate": 5.441888377675536e-06, "loss": 0.0444, "step": 11404 }, { "epoch": 3.54, "learning_rate": 5.441488297659533e-06, "loss": 0.0595, "step": 11405 }, { "epoch": 3.54, "learning_rate": 5.4410882176435284e-06, "loss": 0.0702, "step": 11406 }, { "epoch": 3.54, "learning_rate": 5.440688137627526e-06, "loss": 0.0557, "step": 11407 }, { "epoch": 3.54, "learning_rate": 5.440288057611523e-06, "loss": 0.0259, "step": 11408 }, { "epoch": 3.54, "learning_rate": 5.439887977595519e-06, "loss": 0.0571, "step": 11409 }, { "epoch": 3.54, "learning_rate": 5.439487897579516e-06, "loss": 0.02, "step": 11410 }, { "epoch": 3.54, "learning_rate": 5.439087817563513e-06, "loss": 0.0445, "step": 11411 }, { "epoch": 3.54, "learning_rate": 5.43868773754751e-06, "loss": 0.0432, "step": 11412 }, { "epoch": 3.54, "learning_rate": 5.438287657531507e-06, "loss": 0.0418, "step": 11413 }, { "epoch": 3.54, "learning_rate": 5.437887577515504e-06, "loss": 0.0071, "step": 11414 }, { "epoch": 3.54, "learning_rate": 5.4374874974995e-06, "loss": 0.0284, "step": 11415 }, { "epoch": 3.54, "learning_rate": 5.437087417483498e-06, "loss": 0.1118, "step": 11416 }, { "epoch": 3.54, "learning_rate": 5.4366873374674945e-06, "loss": 0.0847, "step": 11417 }, { "epoch": 3.54, "learning_rate": 5.43628725745149e-06, "loss": 0.0661, "step": 11418 }, { "epoch": 3.54, "learning_rate": 5.435887177435487e-06, "loss": 0.0765, "step": 11419 }, { "epoch": 3.54, "learning_rate": 5.435487097419484e-06, "loss": 0.0454, "step": 11420 }, { "epoch": 3.54, "learning_rate": 5.435087017403481e-06, "loss": 0.0306, "step": 11421 }, { "epoch": 3.55, "learning_rate": 5.434686937387478e-06, "loss": 0.0884, "step": 11422 }, { "epoch": 3.55, "learning_rate": 5.434286857371475e-06, "loss": 0.0312, "step": 11423 }, { "epoch": 3.55, "learning_rate": 5.433886777355471e-06, "loss": 0.0305, "step": 11424 }, { "epoch": 3.55, "learning_rate": 5.433486697339468e-06, "loss": 0.0636, "step": 11425 }, { "epoch": 3.55, "learning_rate": 5.4330866173234655e-06, "loss": 0.0465, "step": 11426 }, { "epoch": 3.55, "learning_rate": 5.432686537307462e-06, "loss": 0.0426, "step": 11427 }, { "epoch": 3.55, "learning_rate": 5.432286457291459e-06, "loss": 0.0457, "step": 11428 }, { "epoch": 3.55, "learning_rate": 5.431886377275456e-06, "loss": 0.0214, "step": 11429 }, { "epoch": 3.55, "learning_rate": 5.431486297259453e-06, "loss": 0.0254, "step": 11430 }, { "epoch": 3.55, "learning_rate": 5.431086217243449e-06, "loss": 0.0333, "step": 11431 }, { "epoch": 3.55, "learning_rate": 5.430686137227446e-06, "loss": 0.0215, "step": 11432 }, { "epoch": 3.55, "learning_rate": 5.430286057211442e-06, "loss": 0.0288, "step": 11433 }, { "epoch": 3.55, "learning_rate": 5.429885977195439e-06, "loss": 0.0055, "step": 11434 }, { "epoch": 3.55, "learning_rate": 5.4294858971794366e-06, "loss": 0.0443, "step": 11435 }, { "epoch": 3.55, "learning_rate": 5.429085817163433e-06, "loss": 0.032, "step": 11436 }, { "epoch": 3.55, "learning_rate": 5.42868573714743e-06, "loss": 0.0516, "step": 11437 }, { "epoch": 3.55, "learning_rate": 5.428285657131427e-06, "loss": 0.0703, "step": 11438 }, { "epoch": 3.55, "learning_rate": 5.427885577115424e-06, "loss": 0.0484, "step": 11439 }, { "epoch": 3.55, "learning_rate": 5.427485497099421e-06, "loss": 0.0567, "step": 11440 }, { "epoch": 3.55, "learning_rate": 5.4270854170834175e-06, "loss": 0.0272, "step": 11441 }, { "epoch": 3.55, "learning_rate": 5.426685337067414e-06, "loss": 0.031, "step": 11442 }, { "epoch": 3.55, "learning_rate": 5.42628525705141e-06, "loss": 0.1249, "step": 11443 }, { "epoch": 3.55, "learning_rate": 5.425885177035407e-06, "loss": 0.0384, "step": 11444 }, { "epoch": 3.55, "learning_rate": 5.425485097019404e-06, "loss": 0.0424, "step": 11445 }, { "epoch": 3.55, "learning_rate": 5.425085017003401e-06, "loss": 0.0313, "step": 11446 }, { "epoch": 3.55, "learning_rate": 5.424684936987398e-06, "loss": 0.0211, "step": 11447 }, { "epoch": 3.55, "learning_rate": 5.424284856971394e-06, "loss": 0.0145, "step": 11448 }, { "epoch": 3.55, "learning_rate": 5.423884776955392e-06, "loss": 0.0237, "step": 11449 }, { "epoch": 3.55, "learning_rate": 5.4234846969393886e-06, "loss": 0.0311, "step": 11450 }, { "epoch": 3.55, "learning_rate": 5.423084616923385e-06, "loss": 0.0337, "step": 11451 }, { "epoch": 3.55, "learning_rate": 5.422684536907382e-06, "loss": 0.0405, "step": 11452 }, { "epoch": 3.55, "learning_rate": 5.4222844568913795e-06, "loss": 0.0397, "step": 11453 }, { "epoch": 3.56, "learning_rate": 5.421884376875376e-06, "loss": 0.0755, "step": 11454 }, { "epoch": 3.56, "learning_rate": 5.421484296859373e-06, "loss": 0.0336, "step": 11455 }, { "epoch": 3.56, "learning_rate": 5.421084216843369e-06, "loss": 0.0741, "step": 11456 }, { "epoch": 3.56, "learning_rate": 5.420684136827365e-06, "loss": 0.06, "step": 11457 }, { "epoch": 3.56, "learning_rate": 5.420284056811362e-06, "loss": 0.0346, "step": 11458 }, { "epoch": 3.56, "learning_rate": 5.41988397679536e-06, "loss": 0.0133, "step": 11459 }, { "epoch": 3.56, "learning_rate": 5.419483896779356e-06, "loss": 0.053, "step": 11460 }, { "epoch": 3.56, "learning_rate": 5.419083816763353e-06, "loss": 0.0677, "step": 11461 }, { "epoch": 3.56, "learning_rate": 5.41868373674735e-06, "loss": 0.0847, "step": 11462 }, { "epoch": 3.56, "learning_rate": 5.418283656731347e-06, "loss": 0.0375, "step": 11463 }, { "epoch": 3.56, "learning_rate": 5.417883576715344e-06, "loss": 0.0569, "step": 11464 }, { "epoch": 3.56, "learning_rate": 5.4174834966993405e-06, "loss": 0.1129, "step": 11465 }, { "epoch": 3.56, "learning_rate": 5.417083416683337e-06, "loss": 0.0334, "step": 11466 }, { "epoch": 3.56, "learning_rate": 5.416683336667335e-06, "loss": 0.028, "step": 11467 }, { "epoch": 3.56, "learning_rate": 5.4162832566513314e-06, "loss": 0.0145, "step": 11468 }, { "epoch": 3.56, "learning_rate": 5.415883176635327e-06, "loss": 0.0499, "step": 11469 }, { "epoch": 3.56, "learning_rate": 5.415483096619324e-06, "loss": 0.0401, "step": 11470 }, { "epoch": 3.56, "learning_rate": 5.415083016603321e-06, "loss": 0.041, "step": 11471 }, { "epoch": 3.56, "learning_rate": 5.414682936587317e-06, "loss": 0.0302, "step": 11472 }, { "epoch": 3.56, "learning_rate": 5.414282856571315e-06, "loss": 0.051, "step": 11473 }, { "epoch": 3.56, "learning_rate": 5.4138827765553116e-06, "loss": 0.0237, "step": 11474 }, { "epoch": 3.56, "learning_rate": 5.413482696539308e-06, "loss": 0.0323, "step": 11475 }, { "epoch": 3.56, "learning_rate": 5.413082616523305e-06, "loss": 0.0439, "step": 11476 }, { "epoch": 3.56, "learning_rate": 5.4126825365073025e-06, "loss": 0.0139, "step": 11477 }, { "epoch": 3.56, "learning_rate": 5.412282456491299e-06, "loss": 0.1108, "step": 11478 }, { "epoch": 3.56, "learning_rate": 5.411882376475296e-06, "loss": 0.027, "step": 11479 }, { "epoch": 3.56, "learning_rate": 5.4114822964592925e-06, "loss": 0.0592, "step": 11480 }, { "epoch": 3.56, "learning_rate": 5.411082216443288e-06, "loss": 0.0418, "step": 11481 }, { "epoch": 3.56, "learning_rate": 5.410682136427285e-06, "loss": 0.0524, "step": 11482 }, { "epoch": 3.56, "learning_rate": 5.410282056411283e-06, "loss": 0.0146, "step": 11483 }, { "epoch": 3.56, "learning_rate": 5.409881976395279e-06, "loss": 0.0217, "step": 11484 }, { "epoch": 3.56, "learning_rate": 5.409481896379276e-06, "loss": 0.0377, "step": 11485 }, { "epoch": 3.57, "learning_rate": 5.409081816363273e-06, "loss": 0.0521, "step": 11486 }, { "epoch": 3.57, "learning_rate": 5.40868173634727e-06, "loss": 0.0412, "step": 11487 }, { "epoch": 3.57, "learning_rate": 5.408281656331267e-06, "loss": 0.0405, "step": 11488 }, { "epoch": 3.57, "learning_rate": 5.4078815763152636e-06, "loss": 0.0407, "step": 11489 }, { "epoch": 3.57, "learning_rate": 5.40748149629926e-06, "loss": 0.1247, "step": 11490 }, { "epoch": 3.57, "learning_rate": 5.407081416283258e-06, "loss": 0.0155, "step": 11491 }, { "epoch": 3.57, "learning_rate": 5.4066813362672545e-06, "loss": 0.0252, "step": 11492 }, { "epoch": 3.57, "learning_rate": 5.406281256251251e-06, "loss": 0.0356, "step": 11493 }, { "epoch": 3.57, "learning_rate": 5.405881176235247e-06, "loss": 0.0658, "step": 11494 }, { "epoch": 3.57, "learning_rate": 5.405481096219244e-06, "loss": 0.0555, "step": 11495 }, { "epoch": 3.57, "learning_rate": 5.40508101620324e-06, "loss": 0.0717, "step": 11496 }, { "epoch": 3.57, "learning_rate": 5.404680936187238e-06, "loss": 0.0401, "step": 11497 }, { "epoch": 3.57, "learning_rate": 5.404280856171235e-06, "loss": 0.022, "step": 11498 }, { "epoch": 3.57, "learning_rate": 5.403880776155231e-06, "loss": 0.0101, "step": 11499 }, { "epoch": 3.57, "learning_rate": 5.403480696139228e-06, "loss": 0.0883, "step": 11500 }, { "epoch": 3.57, "learning_rate": 5.4030806161232255e-06, "loss": 0.0353, "step": 11501 }, { "epoch": 3.57, "learning_rate": 5.402680536107222e-06, "loss": 0.02, "step": 11502 }, { "epoch": 3.57, "learning_rate": 5.402280456091219e-06, "loss": 0.0274, "step": 11503 }, { "epoch": 3.57, "learning_rate": 5.4018803760752155e-06, "loss": 0.0579, "step": 11504 }, { "epoch": 3.57, "learning_rate": 5.401480296059213e-06, "loss": 0.1262, "step": 11505 }, { "epoch": 3.57, "learning_rate": 5.401080216043209e-06, "loss": 0.0396, "step": 11506 }, { "epoch": 3.57, "learning_rate": 5.400680136027206e-06, "loss": 0.0167, "step": 11507 }, { "epoch": 3.57, "learning_rate": 5.400280056011202e-06, "loss": 0.0711, "step": 11508 }, { "epoch": 3.57, "learning_rate": 5.399879975995199e-06, "loss": 0.0392, "step": 11509 }, { "epoch": 3.57, "learning_rate": 5.3994798959791965e-06, "loss": 0.0416, "step": 11510 }, { "epoch": 3.57, "learning_rate": 5.399079815963193e-06, "loss": 0.038, "step": 11511 }, { "epoch": 3.57, "learning_rate": 5.39867973594719e-06, "loss": 0.0636, "step": 11512 }, { "epoch": 3.57, "learning_rate": 5.3982796559311866e-06, "loss": 0.0891, "step": 11513 }, { "epoch": 3.57, "learning_rate": 5.397879575915183e-06, "loss": 0.0274, "step": 11514 }, { "epoch": 3.57, "learning_rate": 5.397479495899181e-06, "loss": 0.046, "step": 11515 }, { "epoch": 3.57, "learning_rate": 5.3970794158831775e-06, "loss": 0.1337, "step": 11516 }, { "epoch": 3.57, "learning_rate": 5.396679335867174e-06, "loss": 0.0195, "step": 11517 }, { "epoch": 3.57, "learning_rate": 5.396279255851171e-06, "loss": 0.0358, "step": 11518 }, { "epoch": 3.58, "learning_rate": 5.395879175835167e-06, "loss": 0.024, "step": 11519 }, { "epoch": 3.58, "learning_rate": 5.395479095819164e-06, "loss": 0.0142, "step": 11520 }, { "epoch": 3.58, "learning_rate": 5.395079015803161e-06, "loss": 0.036, "step": 11521 }, { "epoch": 3.58, "learning_rate": 5.394678935787158e-06, "loss": 0.0568, "step": 11522 }, { "epoch": 3.58, "learning_rate": 5.394278855771154e-06, "loss": 0.0523, "step": 11523 }, { "epoch": 3.58, "learning_rate": 5.393878775755152e-06, "loss": 0.1912, "step": 11524 }, { "epoch": 3.58, "learning_rate": 5.3934786957391485e-06, "loss": 0.0374, "step": 11525 }, { "epoch": 3.58, "learning_rate": 5.393078615723145e-06, "loss": 0.018, "step": 11526 }, { "epoch": 3.58, "learning_rate": 5.392678535707142e-06, "loss": 0.04, "step": 11527 }, { "epoch": 3.58, "learning_rate": 5.3922784556911386e-06, "loss": 0.0533, "step": 11528 }, { "epoch": 3.58, "learning_rate": 5.391878375675136e-06, "loss": 0.0463, "step": 11529 }, { "epoch": 3.58, "learning_rate": 5.391478295659133e-06, "loss": 0.0326, "step": 11530 }, { "epoch": 3.58, "learning_rate": 5.3910782156431295e-06, "loss": 0.0548, "step": 11531 }, { "epoch": 3.58, "learning_rate": 5.390678135627125e-06, "loss": 0.0392, "step": 11532 }, { "epoch": 3.58, "learning_rate": 5.390278055611122e-06, "loss": 0.009, "step": 11533 }, { "epoch": 3.58, "learning_rate": 5.3898779755951195e-06, "loss": 0.0694, "step": 11534 }, { "epoch": 3.58, "learning_rate": 5.389477895579116e-06, "loss": 0.0311, "step": 11535 }, { "epoch": 3.58, "learning_rate": 5.389077815563113e-06, "loss": 0.1432, "step": 11536 }, { "epoch": 3.58, "learning_rate": 5.38867773554711e-06, "loss": 0.0218, "step": 11537 }, { "epoch": 3.58, "learning_rate": 5.388277655531107e-06, "loss": 0.0171, "step": 11538 }, { "epoch": 3.58, "learning_rate": 5.387877575515104e-06, "loss": 0.0704, "step": 11539 }, { "epoch": 3.58, "learning_rate": 5.3874774954991005e-06, "loss": 0.0274, "step": 11540 }, { "epoch": 3.58, "learning_rate": 5.387077415483097e-06, "loss": 0.0873, "step": 11541 }, { "epoch": 3.58, "learning_rate": 5.386677335467094e-06, "loss": 0.0287, "step": 11542 }, { "epoch": 3.58, "learning_rate": 5.386277255451091e-06, "loss": 0.0539, "step": 11543 }, { "epoch": 3.58, "learning_rate": 5.385877175435087e-06, "loss": 0.0347, "step": 11544 }, { "epoch": 3.58, "learning_rate": 5.385477095419084e-06, "loss": 0.0593, "step": 11545 }, { "epoch": 3.58, "learning_rate": 5.385077015403081e-06, "loss": 0.067, "step": 11546 }, { "epoch": 3.58, "learning_rate": 5.384676935387077e-06, "loss": 0.0424, "step": 11547 }, { "epoch": 3.58, "learning_rate": 5.384276855371075e-06, "loss": 0.124, "step": 11548 }, { "epoch": 3.58, "learning_rate": 5.3838767753550715e-06, "loss": 0.0401, "step": 11549 }, { "epoch": 3.58, "learning_rate": 5.383476695339068e-06, "loss": 0.0162, "step": 11550 }, { "epoch": 3.59, "learning_rate": 5.383076615323065e-06, "loss": 0.0195, "step": 11551 }, { "epoch": 3.59, "learning_rate": 5.382676535307062e-06, "loss": 0.0573, "step": 11552 }, { "epoch": 3.59, "learning_rate": 5.382276455291059e-06, "loss": 0.0579, "step": 11553 }, { "epoch": 3.59, "learning_rate": 5.381876375275056e-06, "loss": 0.0675, "step": 11554 }, { "epoch": 3.59, "learning_rate": 5.3814762952590525e-06, "loss": 0.0847, "step": 11555 }, { "epoch": 3.59, "learning_rate": 5.381076215243049e-06, "loss": 0.0355, "step": 11556 }, { "epoch": 3.59, "learning_rate": 5.380676135227045e-06, "loss": 0.1775, "step": 11557 }, { "epoch": 3.59, "learning_rate": 5.3802760552110425e-06, "loss": 0.0443, "step": 11558 }, { "epoch": 3.59, "learning_rate": 5.379875975195039e-06, "loss": 0.1185, "step": 11559 }, { "epoch": 3.59, "learning_rate": 5.379475895179036e-06, "loss": 0.0405, "step": 11560 }, { "epoch": 3.59, "learning_rate": 5.379075815163033e-06, "loss": 0.0292, "step": 11561 }, { "epoch": 3.59, "learning_rate": 5.37867573514703e-06, "loss": 0.0269, "step": 11562 }, { "epoch": 3.59, "learning_rate": 5.378275655131027e-06, "loss": 0.0483, "step": 11563 }, { "epoch": 3.59, "learning_rate": 5.3778755751150235e-06, "loss": 0.042, "step": 11564 }, { "epoch": 3.59, "learning_rate": 5.37747549509902e-06, "loss": 0.034, "step": 11565 }, { "epoch": 3.59, "learning_rate": 5.377075415083018e-06, "loss": 0.1128, "step": 11566 }, { "epoch": 3.59, "learning_rate": 5.376675335067014e-06, "loss": 0.0337, "step": 11567 }, { "epoch": 3.59, "learning_rate": 5.376275255051011e-06, "loss": 0.05, "step": 11568 }, { "epoch": 3.59, "learning_rate": 5.375875175035007e-06, "loss": 0.0422, "step": 11569 }, { "epoch": 3.59, "learning_rate": 5.375475095019004e-06, "loss": 0.088, "step": 11570 }, { "epoch": 3.59, "learning_rate": 5.375075015003e-06, "loss": 0.0365, "step": 11571 }, { "epoch": 3.59, "learning_rate": 5.374674934986998e-06, "loss": 0.0611, "step": 11572 }, { "epoch": 3.59, "learning_rate": 5.3742748549709945e-06, "loss": 0.0429, "step": 11573 }, { "epoch": 3.59, "learning_rate": 5.373874774954991e-06, "loss": 0.0393, "step": 11574 }, { "epoch": 3.59, "learning_rate": 5.373474694938988e-06, "loss": 0.0461, "step": 11575 }, { "epoch": 3.59, "learning_rate": 5.3730746149229854e-06, "loss": 0.0293, "step": 11576 }, { "epoch": 3.59, "learning_rate": 5.372674534906982e-06, "loss": 0.0467, "step": 11577 }, { "epoch": 3.59, "learning_rate": 5.372274454890979e-06, "loss": 0.0106, "step": 11578 }, { "epoch": 3.59, "learning_rate": 5.3718743748749755e-06, "loss": 0.1458, "step": 11579 }, { "epoch": 3.59, "learning_rate": 5.371474294858973e-06, "loss": 0.0522, "step": 11580 }, { "epoch": 3.59, "learning_rate": 5.37107421484297e-06, "loss": 0.0392, "step": 11581 }, { "epoch": 3.59, "learning_rate": 5.3706741348269655e-06, "loss": 0.0462, "step": 11582 }, { "epoch": 3.6, "learning_rate": 5.370274054810962e-06, "loss": 0.0782, "step": 11583 }, { "epoch": 3.6, "learning_rate": 5.369873974794959e-06, "loss": 0.0196, "step": 11584 }, { "epoch": 3.6, "learning_rate": 5.369473894778956e-06, "loss": 0.0486, "step": 11585 }, { "epoch": 3.6, "learning_rate": 5.369073814762953e-06, "loss": 0.0414, "step": 11586 }, { "epoch": 3.6, "learning_rate": 5.36867373474695e-06, "loss": 0.0422, "step": 11587 }, { "epoch": 3.6, "learning_rate": 5.3682736547309465e-06, "loss": 0.0606, "step": 11588 }, { "epoch": 3.6, "learning_rate": 5.367873574714943e-06, "loss": 0.0247, "step": 11589 }, { "epoch": 3.6, "learning_rate": 5.367473494698941e-06, "loss": 0.0211, "step": 11590 }, { "epoch": 3.6, "learning_rate": 5.367073414682937e-06, "loss": 0.0905, "step": 11591 }, { "epoch": 3.6, "learning_rate": 5.366673334666934e-06, "loss": 0.0527, "step": 11592 }, { "epoch": 3.6, "learning_rate": 5.366273254650931e-06, "loss": 0.0227, "step": 11593 }, { "epoch": 3.6, "learning_rate": 5.365873174634928e-06, "loss": 0.0476, "step": 11594 }, { "epoch": 3.6, "learning_rate": 5.365473094618924e-06, "loss": 0.0383, "step": 11595 }, { "epoch": 3.6, "learning_rate": 5.365073014602921e-06, "loss": 0.0159, "step": 11596 }, { "epoch": 3.6, "learning_rate": 5.3646729345869175e-06, "loss": 0.1377, "step": 11597 }, { "epoch": 3.6, "learning_rate": 5.364272854570914e-06, "loss": 0.0192, "step": 11598 }, { "epoch": 3.6, "learning_rate": 5.363872774554911e-06, "loss": 0.025, "step": 11599 }, { "epoch": 3.6, "learning_rate": 5.3634726945389084e-06, "loss": 0.0823, "step": 11600 }, { "epoch": 3.6, "learning_rate": 5.363072614522905e-06, "loss": 0.0325, "step": 11601 }, { "epoch": 3.6, "learning_rate": 5.362672534506902e-06, "loss": 0.0567, "step": 11602 }, { "epoch": 3.6, "learning_rate": 5.3622724544908985e-06, "loss": 0.0537, "step": 11603 }, { "epoch": 3.6, "learning_rate": 5.361872374474896e-06, "loss": 0.0316, "step": 11604 }, { "epoch": 3.6, "learning_rate": 5.361472294458893e-06, "loss": 0.0722, "step": 11605 }, { "epoch": 3.6, "learning_rate": 5.361072214442889e-06, "loss": 0.1684, "step": 11606 }, { "epoch": 3.6, "learning_rate": 5.360672134426885e-06, "loss": 0.0417, "step": 11607 }, { "epoch": 3.6, "learning_rate": 5.360272054410882e-06, "loss": 0.0553, "step": 11608 }, { "epoch": 3.6, "learning_rate": 5.3598719743948795e-06, "loss": 0.0751, "step": 11609 }, { "epoch": 3.6, "learning_rate": 5.359471894378876e-06, "loss": 0.0847, "step": 11610 }, { "epoch": 3.6, "learning_rate": 5.359071814362873e-06, "loss": 0.0251, "step": 11611 }, { "epoch": 3.6, "learning_rate": 5.3586717343468695e-06, "loss": 0.0601, "step": 11612 }, { "epoch": 3.6, "learning_rate": 5.358271654330866e-06, "loss": 0.0951, "step": 11613 }, { "epoch": 3.6, "learning_rate": 5.357871574314864e-06, "loss": 0.1225, "step": 11614 }, { "epoch": 3.61, "learning_rate": 5.3574714942988604e-06, "loss": 0.0377, "step": 11615 }, { "epoch": 3.61, "learning_rate": 5.357071414282857e-06, "loss": 0.0406, "step": 11616 }, { "epoch": 3.61, "learning_rate": 5.356671334266854e-06, "loss": 0.1088, "step": 11617 }, { "epoch": 3.61, "learning_rate": 5.356271254250851e-06, "loss": 0.0354, "step": 11618 }, { "epoch": 3.61, "learning_rate": 5.355871174234848e-06, "loss": 0.0291, "step": 11619 }, { "epoch": 3.61, "learning_rate": 5.355471094218844e-06, "loss": 0.0901, "step": 11620 }, { "epoch": 3.61, "learning_rate": 5.3550710142028405e-06, "loss": 0.0387, "step": 11621 }, { "epoch": 3.61, "learning_rate": 5.354670934186837e-06, "loss": 0.0298, "step": 11622 }, { "epoch": 3.61, "learning_rate": 5.354270854170835e-06, "loss": 0.0235, "step": 11623 }, { "epoch": 3.61, "learning_rate": 5.3538707741548314e-06, "loss": 0.1113, "step": 11624 }, { "epoch": 3.61, "learning_rate": 5.353470694138828e-06, "loss": 0.0472, "step": 11625 }, { "epoch": 3.61, "learning_rate": 5.353070614122825e-06, "loss": 0.0204, "step": 11626 }, { "epoch": 3.61, "learning_rate": 5.3526705341068215e-06, "loss": 0.0396, "step": 11627 }, { "epoch": 3.61, "learning_rate": 5.352270454090819e-06, "loss": 0.0172, "step": 11628 }, { "epoch": 3.61, "learning_rate": 5.351870374074816e-06, "loss": 0.058, "step": 11629 }, { "epoch": 3.61, "learning_rate": 5.351470294058812e-06, "loss": 0.0489, "step": 11630 }, { "epoch": 3.61, "learning_rate": 5.351070214042809e-06, "loss": 0.0451, "step": 11631 }, { "epoch": 3.61, "learning_rate": 5.350670134026807e-06, "loss": 0.0275, "step": 11632 }, { "epoch": 3.61, "learning_rate": 5.3502700540108025e-06, "loss": 0.0761, "step": 11633 }, { "epoch": 3.61, "learning_rate": 5.349869973994799e-06, "loss": 0.0901, "step": 11634 }, { "epoch": 3.61, "learning_rate": 5.349469893978796e-06, "loss": 0.0686, "step": 11635 }, { "epoch": 3.61, "learning_rate": 5.3490698139627925e-06, "loss": 0.0358, "step": 11636 }, { "epoch": 3.61, "learning_rate": 5.34866973394679e-06, "loss": 0.0734, "step": 11637 }, { "epoch": 3.61, "learning_rate": 5.348269653930787e-06, "loss": 0.0687, "step": 11638 }, { "epoch": 3.61, "learning_rate": 5.3478695739147834e-06, "loss": 0.0096, "step": 11639 }, { "epoch": 3.61, "learning_rate": 5.34746949389878e-06, "loss": 0.0442, "step": 11640 }, { "epoch": 3.61, "learning_rate": 5.347069413882777e-06, "loss": 0.0372, "step": 11641 }, { "epoch": 3.61, "learning_rate": 5.346669333866774e-06, "loss": 0.0288, "step": 11642 }, { "epoch": 3.61, "learning_rate": 5.346269253850771e-06, "loss": 0.0298, "step": 11643 }, { "epoch": 3.61, "learning_rate": 5.345869173834768e-06, "loss": 0.0307, "step": 11644 }, { "epoch": 3.61, "learning_rate": 5.3454690938187636e-06, "loss": 0.118, "step": 11645 }, { "epoch": 3.61, "learning_rate": 5.34506901380276e-06, "loss": 0.0484, "step": 11646 }, { "epoch": 3.61, "learning_rate": 5.344668933786758e-06, "loss": 0.038, "step": 11647 }, { "epoch": 3.62, "learning_rate": 5.3442688537707545e-06, "loss": 0.0709, "step": 11648 }, { "epoch": 3.62, "learning_rate": 5.343868773754751e-06, "loss": 0.0328, "step": 11649 }, { "epoch": 3.62, "learning_rate": 5.343468693738748e-06, "loss": 0.0067, "step": 11650 }, { "epoch": 3.62, "learning_rate": 5.343068613722745e-06, "loss": 0.0232, "step": 11651 }, { "epoch": 3.62, "learning_rate": 5.342668533706742e-06, "loss": 0.0376, "step": 11652 }, { "epoch": 3.62, "learning_rate": 5.342268453690739e-06, "loss": 0.0749, "step": 11653 }, { "epoch": 3.62, "learning_rate": 5.3418683736747354e-06, "loss": 0.0268, "step": 11654 }, { "epoch": 3.62, "learning_rate": 5.341468293658733e-06, "loss": 0.0553, "step": 11655 }, { "epoch": 3.62, "learning_rate": 5.34106821364273e-06, "loss": 0.0406, "step": 11656 }, { "epoch": 3.62, "learning_rate": 5.340668133626726e-06, "loss": 0.0611, "step": 11657 }, { "epoch": 3.62, "learning_rate": 5.340268053610722e-06, "loss": 0.0349, "step": 11658 }, { "epoch": 3.62, "learning_rate": 5.339867973594719e-06, "loss": 0.0846, "step": 11659 }, { "epoch": 3.62, "learning_rate": 5.3394678935787155e-06, "loss": 0.055, "step": 11660 }, { "epoch": 3.62, "learning_rate": 5.339067813562713e-06, "loss": 0.1515, "step": 11661 }, { "epoch": 3.62, "learning_rate": 5.33866773354671e-06, "loss": 0.0226, "step": 11662 }, { "epoch": 3.62, "learning_rate": 5.3382676535307064e-06, "loss": 0.064, "step": 11663 }, { "epoch": 3.62, "learning_rate": 5.337867573514703e-06, "loss": 0.0455, "step": 11664 }, { "epoch": 3.62, "learning_rate": 5.337467493498701e-06, "loss": 0.022, "step": 11665 }, { "epoch": 3.62, "learning_rate": 5.337067413482697e-06, "loss": 0.0534, "step": 11666 }, { "epoch": 3.62, "learning_rate": 5.336667333466694e-06, "loss": 0.1709, "step": 11667 }, { "epoch": 3.62, "learning_rate": 5.336267253450691e-06, "loss": 0.0648, "step": 11668 }, { "epoch": 3.62, "learning_rate": 5.335867173434688e-06, "loss": 0.0397, "step": 11669 }, { "epoch": 3.62, "learning_rate": 5.335467093418683e-06, "loss": 0.0177, "step": 11670 }, { "epoch": 3.62, "learning_rate": 5.335067013402681e-06, "loss": 0.0277, "step": 11671 }, { "epoch": 3.62, "learning_rate": 5.3346669333866775e-06, "loss": 0.0392, "step": 11672 }, { "epoch": 3.62, "learning_rate": 5.334266853370674e-06, "loss": 0.0296, "step": 11673 }, { "epoch": 3.62, "learning_rate": 5.333866773354671e-06, "loss": 0.0358, "step": 11674 }, { "epoch": 3.62, "learning_rate": 5.333466693338668e-06, "loss": 0.0361, "step": 11675 }, { "epoch": 3.62, "learning_rate": 5.333066613322665e-06, "loss": 0.0577, "step": 11676 }, { "epoch": 3.62, "learning_rate": 5.332666533306662e-06, "loss": 0.0476, "step": 11677 }, { "epoch": 3.62, "learning_rate": 5.3322664532906584e-06, "loss": 0.0663, "step": 11678 }, { "epoch": 3.62, "learning_rate": 5.331866373274656e-06, "loss": 0.0301, "step": 11679 }, { "epoch": 3.63, "learning_rate": 5.331466293258653e-06, "loss": 0.0394, "step": 11680 }, { "epoch": 3.63, "learning_rate": 5.331066213242649e-06, "loss": 0.1282, "step": 11681 }, { "epoch": 3.63, "learning_rate": 5.330666133226646e-06, "loss": 0.0348, "step": 11682 }, { "epoch": 3.63, "learning_rate": 5.330266053210642e-06, "loss": 0.0319, "step": 11683 }, { "epoch": 3.63, "learning_rate": 5.3298659731946386e-06, "loss": 0.0291, "step": 11684 }, { "epoch": 3.63, "learning_rate": 5.329465893178636e-06, "loss": 0.0359, "step": 11685 }, { "epoch": 3.63, "learning_rate": 5.329065813162633e-06, "loss": 0.0361, "step": 11686 }, { "epoch": 3.63, "learning_rate": 5.3286657331466295e-06, "loss": 0.0242, "step": 11687 }, { "epoch": 3.63, "learning_rate": 5.328265653130626e-06, "loss": 0.124, "step": 11688 }, { "epoch": 3.63, "learning_rate": 5.327865573114624e-06, "loss": 0.0363, "step": 11689 }, { "epoch": 3.63, "learning_rate": 5.32746549309862e-06, "loss": 0.0254, "step": 11690 }, { "epoch": 3.63, "learning_rate": 5.327065413082617e-06, "loss": 0.0111, "step": 11691 }, { "epoch": 3.63, "learning_rate": 5.326665333066614e-06, "loss": 0.0186, "step": 11692 }, { "epoch": 3.63, "learning_rate": 5.326265253050611e-06, "loss": 0.0435, "step": 11693 }, { "epoch": 3.63, "learning_rate": 5.325865173034608e-06, "loss": 0.0291, "step": 11694 }, { "epoch": 3.63, "learning_rate": 5.325465093018605e-06, "loss": 0.0883, "step": 11695 }, { "epoch": 3.63, "learning_rate": 5.3250650130026005e-06, "loss": 0.0656, "step": 11696 }, { "epoch": 3.63, "learning_rate": 5.324664932986597e-06, "loss": 0.0698, "step": 11697 }, { "epoch": 3.63, "learning_rate": 5.324264852970594e-06, "loss": 0.0404, "step": 11698 }, { "epoch": 3.63, "learning_rate": 5.323864772954591e-06, "loss": 0.0695, "step": 11699 }, { "epoch": 3.63, "learning_rate": 5.323464692938588e-06, "loss": 0.0452, "step": 11700 }, { "epoch": 3.63, "learning_rate": 5.323064612922585e-06, "loss": 0.03, "step": 11701 }, { "epoch": 3.63, "learning_rate": 5.3226645329065814e-06, "loss": 0.0544, "step": 11702 }, { "epoch": 3.63, "learning_rate": 5.322264452890579e-06, "loss": 0.0466, "step": 11703 }, { "epoch": 3.63, "learning_rate": 5.321864372874576e-06, "loss": 0.0106, "step": 11704 }, { "epoch": 3.63, "learning_rate": 5.321464292858572e-06, "loss": 0.0394, "step": 11705 }, { "epoch": 3.63, "learning_rate": 5.321064212842569e-06, "loss": 0.029, "step": 11706 }, { "epoch": 3.63, "learning_rate": 5.3206641328265666e-06, "loss": 0.09, "step": 11707 }, { "epoch": 3.63, "learning_rate": 5.320264052810562e-06, "loss": 0.0395, "step": 11708 }, { "epoch": 3.63, "learning_rate": 5.319863972794559e-06, "loss": 0.0858, "step": 11709 }, { "epoch": 3.63, "learning_rate": 5.319463892778556e-06, "loss": 0.0348, "step": 11710 }, { "epoch": 3.63, "learning_rate": 5.3190638127625525e-06, "loss": 0.052, "step": 11711 }, { "epoch": 3.64, "learning_rate": 5.318663732746549e-06, "loss": 0.0263, "step": 11712 }, { "epoch": 3.64, "learning_rate": 5.318263652730547e-06, "loss": 0.1246, "step": 11713 }, { "epoch": 3.64, "learning_rate": 5.317863572714543e-06, "loss": 0.0359, "step": 11714 }, { "epoch": 3.64, "learning_rate": 5.31746349269854e-06, "loss": 0.069, "step": 11715 }, { "epoch": 3.64, "learning_rate": 5.317063412682537e-06, "loss": 0.0319, "step": 11716 }, { "epoch": 3.64, "learning_rate": 5.316663332666534e-06, "loss": 0.0379, "step": 11717 }, { "epoch": 3.64, "learning_rate": 5.316263252650531e-06, "loss": 0.0386, "step": 11718 }, { "epoch": 3.64, "learning_rate": 5.315863172634528e-06, "loss": 0.0597, "step": 11719 }, { "epoch": 3.64, "learning_rate": 5.315463092618524e-06, "loss": 0.0324, "step": 11720 }, { "epoch": 3.64, "learning_rate": 5.31506301260252e-06, "loss": 0.1042, "step": 11721 }, { "epoch": 3.64, "learning_rate": 5.314662932586518e-06, "loss": 0.0463, "step": 11722 }, { "epoch": 3.64, "learning_rate": 5.314262852570514e-06, "loss": 0.027, "step": 11723 }, { "epoch": 3.64, "learning_rate": 5.313862772554511e-06, "loss": 0.0551, "step": 11724 }, { "epoch": 3.64, "learning_rate": 5.313462692538508e-06, "loss": 0.043, "step": 11725 }, { "epoch": 3.64, "learning_rate": 5.3130626125225045e-06, "loss": 0.0216, "step": 11726 }, { "epoch": 3.64, "learning_rate": 5.312662532506502e-06, "loss": 0.0283, "step": 11727 }, { "epoch": 3.64, "learning_rate": 5.312262452490499e-06, "loss": 0.0403, "step": 11728 }, { "epoch": 3.64, "learning_rate": 5.311862372474495e-06, "loss": 0.0238, "step": 11729 }, { "epoch": 3.64, "learning_rate": 5.311462292458492e-06, "loss": 0.0761, "step": 11730 }, { "epoch": 3.64, "learning_rate": 5.31106221244249e-06, "loss": 0.0334, "step": 11731 }, { "epoch": 3.64, "learning_rate": 5.310662132426486e-06, "loss": 0.0413, "step": 11732 }, { "epoch": 3.64, "learning_rate": 5.310262052410482e-06, "loss": 0.0451, "step": 11733 }, { "epoch": 3.64, "learning_rate": 5.309861972394479e-06, "loss": 0.0632, "step": 11734 }, { "epoch": 3.64, "learning_rate": 5.3094618923784755e-06, "loss": 0.0428, "step": 11735 }, { "epoch": 3.64, "learning_rate": 5.309061812362473e-06, "loss": 0.0508, "step": 11736 }, { "epoch": 3.64, "learning_rate": 5.30866173234647e-06, "loss": 0.0291, "step": 11737 }, { "epoch": 3.64, "learning_rate": 5.308261652330466e-06, "loss": 0.0201, "step": 11738 }, { "epoch": 3.64, "learning_rate": 5.307861572314463e-06, "loss": 0.0544, "step": 11739 }, { "epoch": 3.64, "learning_rate": 5.307461492298461e-06, "loss": 0.0386, "step": 11740 }, { "epoch": 3.64, "learning_rate": 5.307061412282457e-06, "loss": 0.1018, "step": 11741 }, { "epoch": 3.64, "learning_rate": 5.306661332266454e-06, "loss": 0.0345, "step": 11742 }, { "epoch": 3.64, "learning_rate": 5.306261252250451e-06, "loss": 0.2036, "step": 11743 }, { "epoch": 3.65, "learning_rate": 5.305861172234447e-06, "loss": 0.0298, "step": 11744 }, { "epoch": 3.65, "learning_rate": 5.305461092218445e-06, "loss": 0.1227, "step": 11745 }, { "epoch": 3.65, "learning_rate": 5.305061012202441e-06, "loss": 0.033, "step": 11746 }, { "epoch": 3.65, "learning_rate": 5.304660932186437e-06, "loss": 0.0276, "step": 11747 }, { "epoch": 3.65, "learning_rate": 5.304260852170434e-06, "loss": 0.0405, "step": 11748 }, { "epoch": 3.65, "learning_rate": 5.303860772154431e-06, "loss": 0.0418, "step": 11749 }, { "epoch": 3.65, "learning_rate": 5.303460692138428e-06, "loss": 0.0736, "step": 11750 }, { "epoch": 3.65, "learning_rate": 5.303060612122425e-06, "loss": 0.036, "step": 11751 }, { "epoch": 3.65, "learning_rate": 5.302660532106422e-06, "loss": 0.0361, "step": 11752 }, { "epoch": 3.65, "learning_rate": 5.302260452090418e-06, "loss": 0.028, "step": 11753 }, { "epoch": 3.65, "learning_rate": 5.301860372074416e-06, "loss": 0.0213, "step": 11754 }, { "epoch": 3.65, "learning_rate": 5.301460292058413e-06, "loss": 0.0461, "step": 11755 }, { "epoch": 3.65, "learning_rate": 5.301060212042409e-06, "loss": 0.0986, "step": 11756 }, { "epoch": 3.65, "learning_rate": 5.300660132026406e-06, "loss": 0.0758, "step": 11757 }, { "epoch": 3.65, "learning_rate": 5.300260052010403e-06, "loss": 0.04, "step": 11758 }, { "epoch": 3.65, "learning_rate": 5.2998599719943985e-06, "loss": 0.1334, "step": 11759 }, { "epoch": 3.65, "learning_rate": 5.299459891978396e-06, "loss": 0.0429, "step": 11760 }, { "epoch": 3.65, "learning_rate": 5.299059811962393e-06, "loss": 0.0481, "step": 11761 }, { "epoch": 3.65, "learning_rate": 5.298659731946389e-06, "loss": 0.0413, "step": 11762 }, { "epoch": 3.65, "learning_rate": 5.298259651930386e-06, "loss": 0.0488, "step": 11763 }, { "epoch": 3.65, "learning_rate": 5.297859571914384e-06, "loss": 0.0377, "step": 11764 }, { "epoch": 3.65, "learning_rate": 5.29745949189838e-06, "loss": 0.0343, "step": 11765 }, { "epoch": 3.65, "learning_rate": 5.297059411882377e-06, "loss": 0.0283, "step": 11766 }, { "epoch": 3.65, "learning_rate": 5.296659331866374e-06, "loss": 0.0506, "step": 11767 }, { "epoch": 3.65, "learning_rate": 5.296259251850371e-06, "loss": 0.056, "step": 11768 }, { "epoch": 3.65, "learning_rate": 5.295859171834368e-06, "loss": 0.0641, "step": 11769 }, { "epoch": 3.65, "learning_rate": 5.295459091818365e-06, "loss": 0.0537, "step": 11770 }, { "epoch": 3.65, "learning_rate": 5.2950590118023604e-06, "loss": 0.0566, "step": 11771 }, { "epoch": 3.65, "learning_rate": 5.294658931786357e-06, "loss": 0.0507, "step": 11772 }, { "epoch": 3.65, "learning_rate": 5.294258851770354e-06, "loss": 0.0898, "step": 11773 }, { "epoch": 3.65, "learning_rate": 5.293858771754351e-06, "loss": 0.0789, "step": 11774 }, { "epoch": 3.65, "learning_rate": 5.293458691738348e-06, "loss": 0.1257, "step": 11775 }, { "epoch": 3.66, "learning_rate": 5.293058611722345e-06, "loss": 0.0194, "step": 11776 }, { "epoch": 3.66, "learning_rate": 5.292658531706341e-06, "loss": 0.0412, "step": 11777 }, { "epoch": 3.66, "learning_rate": 5.292258451690339e-06, "loss": 0.0539, "step": 11778 }, { "epoch": 3.66, "learning_rate": 5.291858371674336e-06, "loss": 0.0222, "step": 11779 }, { "epoch": 3.66, "learning_rate": 5.291458291658332e-06, "loss": 0.093, "step": 11780 }, { "epoch": 3.66, "learning_rate": 5.291058211642329e-06, "loss": 0.0279, "step": 11781 }, { "epoch": 3.66, "learning_rate": 5.2906581316263265e-06, "loss": 0.0273, "step": 11782 }, { "epoch": 3.66, "learning_rate": 5.290258051610323e-06, "loss": 0.0705, "step": 11783 }, { "epoch": 3.66, "learning_rate": 5.289857971594319e-06, "loss": 0.0508, "step": 11784 }, { "epoch": 3.66, "learning_rate": 5.289457891578316e-06, "loss": 0.0187, "step": 11785 }, { "epoch": 3.66, "learning_rate": 5.289057811562312e-06, "loss": 0.0651, "step": 11786 }, { "epoch": 3.66, "learning_rate": 5.288657731546309e-06, "loss": 0.0363, "step": 11787 }, { "epoch": 3.66, "learning_rate": 5.288257651530307e-06, "loss": 0.0525, "step": 11788 }, { "epoch": 3.66, "learning_rate": 5.287857571514303e-06, "loss": 0.0195, "step": 11789 }, { "epoch": 3.66, "learning_rate": 5.2874574914983e-06, "loss": 0.0358, "step": 11790 }, { "epoch": 3.66, "learning_rate": 5.287057411482297e-06, "loss": 0.0447, "step": 11791 }, { "epoch": 3.66, "learning_rate": 5.286657331466294e-06, "loss": 0.0293, "step": 11792 }, { "epoch": 3.66, "learning_rate": 5.286257251450291e-06, "loss": 0.0129, "step": 11793 }, { "epoch": 3.66, "learning_rate": 5.285857171434288e-06, "loss": 0.0743, "step": 11794 }, { "epoch": 3.66, "learning_rate": 5.285457091418284e-06, "loss": 0.0294, "step": 11795 }, { "epoch": 3.66, "learning_rate": 5.28505701140228e-06, "loss": 0.0597, "step": 11796 }, { "epoch": 3.66, "learning_rate": 5.284656931386277e-06, "loss": 0.0247, "step": 11797 }, { "epoch": 3.66, "learning_rate": 5.284256851370274e-06, "loss": 0.0256, "step": 11798 }, { "epoch": 3.66, "learning_rate": 5.283856771354271e-06, "loss": 0.0335, "step": 11799 }, { "epoch": 3.66, "learning_rate": 5.283456691338268e-06, "loss": 0.019, "step": 11800 }, { "epoch": 3.66, "learning_rate": 5.283056611322264e-06, "loss": 0.0354, "step": 11801 }, { "epoch": 3.66, "learning_rate": 5.282656531306262e-06, "loss": 0.0162, "step": 11802 }, { "epoch": 3.66, "learning_rate": 5.282256451290259e-06, "loss": 0.0479, "step": 11803 }, { "epoch": 3.66, "learning_rate": 5.281856371274255e-06, "loss": 0.0517, "step": 11804 }, { "epoch": 3.66, "learning_rate": 5.281456291258252e-06, "loss": 0.0669, "step": 11805 }, { "epoch": 3.66, "learning_rate": 5.2810562112422495e-06, "loss": 0.039, "step": 11806 }, { "epoch": 3.66, "learning_rate": 5.280656131226246e-06, "loss": 0.0359, "step": 11807 }, { "epoch": 3.66, "learning_rate": 5.280256051210243e-06, "loss": 0.0264, "step": 11808 }, { "epoch": 3.67, "learning_rate": 5.279855971194239e-06, "loss": 0.0537, "step": 11809 }, { "epoch": 3.67, "learning_rate": 5.2794558911782354e-06, "loss": 0.0719, "step": 11810 }, { "epoch": 3.67, "learning_rate": 5.279055811162233e-06, "loss": 0.0672, "step": 11811 }, { "epoch": 3.67, "learning_rate": 5.27865573114623e-06, "loss": 0.0895, "step": 11812 }, { "epoch": 3.67, "learning_rate": 5.278255651130226e-06, "loss": 0.11, "step": 11813 }, { "epoch": 3.67, "learning_rate": 5.277855571114223e-06, "loss": 0.0389, "step": 11814 }, { "epoch": 3.67, "learning_rate": 5.27745549109822e-06, "loss": 0.0124, "step": 11815 }, { "epoch": 3.67, "learning_rate": 5.277055411082217e-06, "loss": 0.0518, "step": 11816 }, { "epoch": 3.67, "learning_rate": 5.276655331066214e-06, "loss": 0.0945, "step": 11817 }, { "epoch": 3.67, "learning_rate": 5.276255251050211e-06, "loss": 0.0463, "step": 11818 }, { "epoch": 3.67, "learning_rate": 5.275855171034207e-06, "loss": 0.0319, "step": 11819 }, { "epoch": 3.67, "learning_rate": 5.275455091018205e-06, "loss": 0.0439, "step": 11820 }, { "epoch": 3.67, "learning_rate": 5.2750550110022015e-06, "loss": 0.0368, "step": 11821 }, { "epoch": 3.67, "learning_rate": 5.274654930986197e-06, "loss": 0.106, "step": 11822 }, { "epoch": 3.67, "learning_rate": 5.274254850970194e-06, "loss": 0.0363, "step": 11823 }, { "epoch": 3.67, "learning_rate": 5.273854770954191e-06, "loss": 0.0375, "step": 11824 }, { "epoch": 3.67, "learning_rate": 5.273454690938188e-06, "loss": 0.0556, "step": 11825 }, { "epoch": 3.67, "learning_rate": 5.273054610922185e-06, "loss": 0.0475, "step": 11826 }, { "epoch": 3.67, "learning_rate": 5.272654530906182e-06, "loss": 0.0364, "step": 11827 }, { "epoch": 3.67, "learning_rate": 5.272254450890178e-06, "loss": 0.1475, "step": 11828 }, { "epoch": 3.67, "learning_rate": 5.271854370874175e-06, "loss": 0.0462, "step": 11829 }, { "epoch": 3.67, "learning_rate": 5.2714542908581725e-06, "loss": 0.0283, "step": 11830 }, { "epoch": 3.67, "learning_rate": 5.271054210842169e-06, "loss": 0.0217, "step": 11831 }, { "epoch": 3.67, "learning_rate": 5.270654130826166e-06, "loss": 0.0533, "step": 11832 }, { "epoch": 3.67, "learning_rate": 5.270254050810163e-06, "loss": 0.0447, "step": 11833 }, { "epoch": 3.67, "learning_rate": 5.2698539707941584e-06, "loss": 0.0399, "step": 11834 }, { "epoch": 3.67, "learning_rate": 5.269453890778156e-06, "loss": 0.1456, "step": 11835 }, { "epoch": 3.67, "learning_rate": 5.269053810762153e-06, "loss": 0.0298, "step": 11836 }, { "epoch": 3.67, "learning_rate": 5.268653730746149e-06, "loss": 0.0361, "step": 11837 }, { "epoch": 3.67, "learning_rate": 5.268253650730146e-06, "loss": 0.0505, "step": 11838 }, { "epoch": 3.67, "learning_rate": 5.2678535707141436e-06, "loss": 0.0193, "step": 11839 }, { "epoch": 3.67, "learning_rate": 5.26745349069814e-06, "loss": 0.0208, "step": 11840 }, { "epoch": 3.68, "learning_rate": 5.267053410682137e-06, "loss": 0.0432, "step": 11841 }, { "epoch": 3.68, "learning_rate": 5.266653330666134e-06, "loss": 0.0285, "step": 11842 }, { "epoch": 3.68, "learning_rate": 5.26625325065013e-06, "loss": 0.1892, "step": 11843 }, { "epoch": 3.68, "learning_rate": 5.265853170634128e-06, "loss": 0.0256, "step": 11844 }, { "epoch": 3.68, "learning_rate": 5.2654530906181245e-06, "loss": 0.1662, "step": 11845 }, { "epoch": 3.68, "learning_rate": 5.265053010602121e-06, "loss": 0.0266, "step": 11846 }, { "epoch": 3.68, "learning_rate": 5.264652930586117e-06, "loss": 0.0301, "step": 11847 }, { "epoch": 3.68, "learning_rate": 5.264252850570114e-06, "loss": 0.0382, "step": 11848 }, { "epoch": 3.68, "learning_rate": 5.263852770554111e-06, "loss": 0.0681, "step": 11849 }, { "epoch": 3.68, "learning_rate": 5.263452690538108e-06, "loss": 0.0215, "step": 11850 }, { "epoch": 3.68, "learning_rate": 5.263052610522105e-06, "loss": 0.0426, "step": 11851 }, { "epoch": 3.68, "learning_rate": 5.262652530506101e-06, "loss": 0.0332, "step": 11852 }, { "epoch": 3.68, "learning_rate": 5.262252450490099e-06, "loss": 0.0512, "step": 11853 }, { "epoch": 3.68, "learning_rate": 5.2618523704740955e-06, "loss": 0.0686, "step": 11854 }, { "epoch": 3.68, "learning_rate": 5.261452290458092e-06, "loss": 0.0577, "step": 11855 }, { "epoch": 3.68, "learning_rate": 5.261052210442089e-06, "loss": 0.0182, "step": 11856 }, { "epoch": 3.68, "learning_rate": 5.260652130426086e-06, "loss": 0.0335, "step": 11857 }, { "epoch": 3.68, "learning_rate": 5.260252050410083e-06, "loss": 0.1411, "step": 11858 }, { "epoch": 3.68, "learning_rate": 5.25985197039408e-06, "loss": 0.0491, "step": 11859 }, { "epoch": 3.68, "learning_rate": 5.259451890378076e-06, "loss": 0.0333, "step": 11860 }, { "epoch": 3.68, "learning_rate": 5.259051810362072e-06, "loss": 0.087, "step": 11861 }, { "epoch": 3.68, "learning_rate": 5.258651730346069e-06, "loss": 0.0272, "step": 11862 }, { "epoch": 3.68, "learning_rate": 5.2582516503300666e-06, "loss": 0.0413, "step": 11863 }, { "epoch": 3.68, "learning_rate": 5.257851570314063e-06, "loss": 0.0238, "step": 11864 }, { "epoch": 3.68, "learning_rate": 5.25745149029806e-06, "loss": 0.0361, "step": 11865 }, { "epoch": 3.68, "learning_rate": 5.257051410282057e-06, "loss": 0.0288, "step": 11866 }, { "epoch": 3.68, "learning_rate": 5.256651330266054e-06, "loss": 0.0537, "step": 11867 }, { "epoch": 3.68, "learning_rate": 5.256251250250051e-06, "loss": 0.0543, "step": 11868 }, { "epoch": 3.68, "learning_rate": 5.2558511702340475e-06, "loss": 0.0333, "step": 11869 }, { "epoch": 3.68, "learning_rate": 5.255451090218044e-06, "loss": 0.0444, "step": 11870 }, { "epoch": 3.68, "learning_rate": 5.255051010202042e-06, "loss": 0.0149, "step": 11871 }, { "epoch": 3.68, "learning_rate": 5.254650930186037e-06, "loss": 0.065, "step": 11872 }, { "epoch": 3.69, "learning_rate": 5.254250850170034e-06, "loss": 0.0188, "step": 11873 }, { "epoch": 3.69, "learning_rate": 5.253850770154031e-06, "loss": 0.0226, "step": 11874 }, { "epoch": 3.69, "learning_rate": 5.253450690138028e-06, "loss": 0.0237, "step": 11875 }, { "epoch": 3.69, "learning_rate": 5.253050610122024e-06, "loss": 0.0684, "step": 11876 }, { "epoch": 3.69, "learning_rate": 5.252650530106022e-06, "loss": 0.0473, "step": 11877 }, { "epoch": 3.69, "learning_rate": 5.2522504500900186e-06, "loss": 0.0233, "step": 11878 }, { "epoch": 3.69, "learning_rate": 5.251850370074015e-06, "loss": 0.0234, "step": 11879 }, { "epoch": 3.69, "learning_rate": 5.251450290058012e-06, "loss": 0.0341, "step": 11880 }, { "epoch": 3.69, "learning_rate": 5.2510502100420095e-06, "loss": 0.0308, "step": 11881 }, { "epoch": 3.69, "learning_rate": 5.250650130026006e-06, "loss": 0.0294, "step": 11882 }, { "epoch": 3.69, "learning_rate": 5.250250050010003e-06, "loss": 0.0268, "step": 11883 }, { "epoch": 3.69, "learning_rate": 5.2498499699939995e-06, "loss": 0.0218, "step": 11884 }, { "epoch": 3.69, "learning_rate": 5.249449889977995e-06, "loss": 0.0154, "step": 11885 }, { "epoch": 3.69, "learning_rate": 5.249049809961992e-06, "loss": 0.1266, "step": 11886 }, { "epoch": 3.69, "learning_rate": 5.24864972994599e-06, "loss": 0.0381, "step": 11887 }, { "epoch": 3.69, "learning_rate": 5.248249649929986e-06, "loss": 0.125, "step": 11888 }, { "epoch": 3.69, "learning_rate": 5.247849569913983e-06, "loss": 0.1631, "step": 11889 }, { "epoch": 3.69, "learning_rate": 5.24744948989798e-06, "loss": 0.0427, "step": 11890 }, { "epoch": 3.69, "learning_rate": 5.247049409881977e-06, "loss": 0.037, "step": 11891 }, { "epoch": 3.69, "learning_rate": 5.246649329865974e-06, "loss": 0.0245, "step": 11892 }, { "epoch": 3.69, "learning_rate": 5.2462492498499705e-06, "loss": 0.063, "step": 11893 }, { "epoch": 3.69, "learning_rate": 5.245849169833967e-06, "loss": 0.033, "step": 11894 }, { "epoch": 3.69, "learning_rate": 5.245449089817965e-06, "loss": 0.0356, "step": 11895 }, { "epoch": 3.69, "learning_rate": 5.2450490098019615e-06, "loss": 0.0197, "step": 11896 }, { "epoch": 3.69, "learning_rate": 5.244648929785957e-06, "loss": 0.0875, "step": 11897 }, { "epoch": 3.69, "learning_rate": 5.244248849769954e-06, "loss": 0.0396, "step": 11898 }, { "epoch": 3.69, "learning_rate": 5.243848769753951e-06, "loss": 0.056, "step": 11899 }, { "epoch": 3.69, "learning_rate": 5.243448689737947e-06, "loss": 0.0185, "step": 11900 }, { "epoch": 3.69, "learning_rate": 5.243048609721945e-06, "loss": 0.0214, "step": 11901 }, { "epoch": 3.69, "learning_rate": 5.2426485297059416e-06, "loss": 0.0372, "step": 11902 }, { "epoch": 3.69, "learning_rate": 5.242248449689938e-06, "loss": 0.1348, "step": 11903 }, { "epoch": 3.69, "learning_rate": 5.241848369673935e-06, "loss": 0.0257, "step": 11904 }, { "epoch": 3.7, "learning_rate": 5.2414482896579325e-06, "loss": 0.0396, "step": 11905 }, { "epoch": 3.7, "learning_rate": 5.241048209641929e-06, "loss": 0.0243, "step": 11906 }, { "epoch": 3.7, "learning_rate": 5.240648129625926e-06, "loss": 0.0284, "step": 11907 }, { "epoch": 3.7, "learning_rate": 5.2402480496099225e-06, "loss": 0.0295, "step": 11908 }, { "epoch": 3.7, "learning_rate": 5.23984796959392e-06, "loss": 0.0459, "step": 11909 }, { "epoch": 3.7, "learning_rate": 5.239447889577916e-06, "loss": 0.0161, "step": 11910 }, { "epoch": 3.7, "learning_rate": 5.239047809561913e-06, "loss": 0.0924, "step": 11911 }, { "epoch": 3.7, "learning_rate": 5.238647729545909e-06, "loss": 0.0197, "step": 11912 }, { "epoch": 3.7, "learning_rate": 5.238247649529906e-06, "loss": 0.0228, "step": 11913 }, { "epoch": 3.7, "learning_rate": 5.237847569513903e-06, "loss": 0.0205, "step": 11914 }, { "epoch": 3.7, "learning_rate": 5.2374474894979e-06, "loss": 0.0399, "step": 11915 }, { "epoch": 3.7, "learning_rate": 5.237047409481897e-06, "loss": 0.0381, "step": 11916 }, { "epoch": 3.7, "learning_rate": 5.2366473294658936e-06, "loss": 0.0502, "step": 11917 }, { "epoch": 3.7, "learning_rate": 5.23624724944989e-06, "loss": 0.0214, "step": 11918 }, { "epoch": 3.7, "learning_rate": 5.235847169433888e-06, "loss": 0.0338, "step": 11919 }, { "epoch": 3.7, "learning_rate": 5.2354470894178845e-06, "loss": 0.0651, "step": 11920 }, { "epoch": 3.7, "learning_rate": 5.235047009401881e-06, "loss": 0.0522, "step": 11921 }, { "epoch": 3.7, "learning_rate": 5.234646929385878e-06, "loss": 0.0357, "step": 11922 }, { "epoch": 3.7, "learning_rate": 5.234246849369874e-06, "loss": 0.0596, "step": 11923 }, { "epoch": 3.7, "learning_rate": 5.233846769353871e-06, "loss": 0.0245, "step": 11924 }, { "epoch": 3.7, "learning_rate": 5.233446689337868e-06, "loss": 0.0374, "step": 11925 }, { "epoch": 3.7, "learning_rate": 5.233046609321865e-06, "loss": 0.0577, "step": 11926 }, { "epoch": 3.7, "learning_rate": 5.232646529305861e-06, "loss": 0.083, "step": 11927 }, { "epoch": 3.7, "learning_rate": 5.232246449289858e-06, "loss": 0.1142, "step": 11928 }, { "epoch": 3.7, "learning_rate": 5.2318463692738555e-06, "loss": 0.1379, "step": 11929 }, { "epoch": 3.7, "learning_rate": 5.231446289257852e-06, "loss": 0.0238, "step": 11930 }, { "epoch": 3.7, "learning_rate": 5.231046209241849e-06, "loss": 0.1048, "step": 11931 }, { "epoch": 3.7, "learning_rate": 5.2306461292258455e-06, "loss": 0.0272, "step": 11932 }, { "epoch": 3.7, "learning_rate": 5.230246049209843e-06, "loss": 0.03, "step": 11933 }, { "epoch": 3.7, "learning_rate": 5.22984596919384e-06, "loss": 0.0134, "step": 11934 }, { "epoch": 3.7, "learning_rate": 5.229445889177836e-06, "loss": 0.0783, "step": 11935 }, { "epoch": 3.7, "learning_rate": 5.229045809161832e-06, "loss": 0.0181, "step": 11936 }, { "epoch": 3.7, "learning_rate": 5.228645729145829e-06, "loss": 0.1267, "step": 11937 }, { "epoch": 3.71, "learning_rate": 5.2282456491298265e-06, "loss": 0.0701, "step": 11938 }, { "epoch": 3.71, "learning_rate": 5.227845569113823e-06, "loss": 0.067, "step": 11939 }, { "epoch": 3.71, "learning_rate": 5.22744548909782e-06, "loss": 0.0551, "step": 11940 }, { "epoch": 3.71, "learning_rate": 5.2270454090818166e-06, "loss": 0.0186, "step": 11941 }, { "epoch": 3.71, "learning_rate": 5.226645329065813e-06, "loss": 0.0358, "step": 11942 }, { "epoch": 3.71, "learning_rate": 5.226245249049811e-06, "loss": 0.0339, "step": 11943 }, { "epoch": 3.71, "learning_rate": 5.2258451690338075e-06, "loss": 0.05, "step": 11944 }, { "epoch": 3.71, "learning_rate": 5.225445089017804e-06, "loss": 0.0492, "step": 11945 }, { "epoch": 3.71, "learning_rate": 5.225045009001801e-06, "loss": 0.0283, "step": 11946 }, { "epoch": 3.71, "learning_rate": 5.224644928985798e-06, "loss": 0.0335, "step": 11947 }, { "epoch": 3.71, "learning_rate": 5.224244848969794e-06, "loss": 0.0375, "step": 11948 }, { "epoch": 3.71, "learning_rate": 5.223844768953791e-06, "loss": 0.0332, "step": 11949 }, { "epoch": 3.71, "learning_rate": 5.223444688937788e-06, "loss": 0.0322, "step": 11950 }, { "epoch": 3.71, "learning_rate": 5.223044608921784e-06, "loss": 0.0585, "step": 11951 }, { "epoch": 3.71, "learning_rate": 5.222644528905782e-06, "loss": 0.0323, "step": 11952 }, { "epoch": 3.71, "learning_rate": 5.2222444488897785e-06, "loss": 0.0663, "step": 11953 }, { "epoch": 3.71, "learning_rate": 5.221844368873775e-06, "loss": 0.0492, "step": 11954 }, { "epoch": 3.71, "learning_rate": 5.221444288857772e-06, "loss": 0.0246, "step": 11955 }, { "epoch": 3.71, "learning_rate": 5.221044208841769e-06, "loss": 0.0407, "step": 11956 }, { "epoch": 3.71, "learning_rate": 5.220644128825766e-06, "loss": 0.06, "step": 11957 }, { "epoch": 3.71, "learning_rate": 5.220244048809763e-06, "loss": 0.0999, "step": 11958 }, { "epoch": 3.71, "learning_rate": 5.2198439687937595e-06, "loss": 0.0599, "step": 11959 }, { "epoch": 3.71, "learning_rate": 5.219443888777755e-06, "loss": 0.0103, "step": 11960 }, { "epoch": 3.71, "learning_rate": 5.219043808761752e-06, "loss": 0.0574, "step": 11961 }, { "epoch": 3.71, "learning_rate": 5.2186437287457495e-06, "loss": 0.1275, "step": 11962 }, { "epoch": 3.71, "learning_rate": 5.218243648729746e-06, "loss": 0.0274, "step": 11963 }, { "epoch": 3.71, "learning_rate": 5.217843568713743e-06, "loss": 0.0104, "step": 11964 }, { "epoch": 3.71, "learning_rate": 5.21744348869774e-06, "loss": 0.0439, "step": 11965 }, { "epoch": 3.71, "learning_rate": 5.217043408681737e-06, "loss": 0.0623, "step": 11966 }, { "epoch": 3.71, "learning_rate": 5.216643328665734e-06, "loss": 0.0636, "step": 11967 }, { "epoch": 3.71, "learning_rate": 5.2162432486497305e-06, "loss": 0.0573, "step": 11968 }, { "epoch": 3.71, "learning_rate": 5.215843168633727e-06, "loss": 0.0587, "step": 11969 }, { "epoch": 3.72, "learning_rate": 5.215443088617725e-06, "loss": 0.0819, "step": 11970 }, { "epoch": 3.72, "learning_rate": 5.215043008601721e-06, "loss": 0.0475, "step": 11971 }, { "epoch": 3.72, "learning_rate": 5.214642928585718e-06, "loss": 0.0682, "step": 11972 }, { "epoch": 3.72, "learning_rate": 5.214242848569714e-06, "loss": 0.0682, "step": 11973 }, { "epoch": 3.72, "learning_rate": 5.213842768553711e-06, "loss": 0.0421, "step": 11974 }, { "epoch": 3.72, "learning_rate": 5.213442688537707e-06, "loss": 0.0277, "step": 11975 }, { "epoch": 3.72, "learning_rate": 5.213042608521705e-06, "loss": 0.0301, "step": 11976 }, { "epoch": 3.72, "learning_rate": 5.2126425285057015e-06, "loss": 0.0723, "step": 11977 }, { "epoch": 3.72, "learning_rate": 5.212242448489698e-06, "loss": 0.0477, "step": 11978 }, { "epoch": 3.72, "learning_rate": 5.211842368473695e-06, "loss": 0.0125, "step": 11979 }, { "epoch": 3.72, "learning_rate": 5.211442288457692e-06, "loss": 0.0242, "step": 11980 }, { "epoch": 3.72, "learning_rate": 5.211042208441689e-06, "loss": 0.015, "step": 11981 }, { "epoch": 3.72, "learning_rate": 5.210642128425686e-06, "loss": 0.0234, "step": 11982 }, { "epoch": 3.72, "learning_rate": 5.2102420484096825e-06, "loss": 0.0493, "step": 11983 }, { "epoch": 3.72, "learning_rate": 5.20984196839368e-06, "loss": 0.0763, "step": 11984 }, { "epoch": 3.72, "learning_rate": 5.209441888377677e-06, "loss": 0.0436, "step": 11985 }, { "epoch": 3.72, "learning_rate": 5.2090418083616725e-06, "loss": 0.0497, "step": 11986 }, { "epoch": 3.72, "learning_rate": 5.208641728345669e-06, "loss": 0.0538, "step": 11987 }, { "epoch": 3.72, "learning_rate": 5.208241648329666e-06, "loss": 0.0113, "step": 11988 }, { "epoch": 3.72, "learning_rate": 5.207841568313663e-06, "loss": 0.028, "step": 11989 }, { "epoch": 3.72, "learning_rate": 5.20744148829766e-06, "loss": 0.0228, "step": 11990 }, { "epoch": 3.72, "learning_rate": 5.207041408281657e-06, "loss": 0.0381, "step": 11991 }, { "epoch": 3.72, "learning_rate": 5.2066413282656535e-06, "loss": 0.0645, "step": 11992 }, { "epoch": 3.72, "learning_rate": 5.20624124824965e-06, "loss": 0.0299, "step": 11993 }, { "epoch": 3.72, "learning_rate": 5.205841168233648e-06, "loss": 0.0709, "step": 11994 }, { "epoch": 3.72, "learning_rate": 5.205441088217644e-06, "loss": 0.0319, "step": 11995 }, { "epoch": 3.72, "learning_rate": 5.205041008201641e-06, "loss": 0.0605, "step": 11996 }, { "epoch": 3.72, "learning_rate": 5.204640928185638e-06, "loss": 0.1068, "step": 11997 }, { "epoch": 3.72, "learning_rate": 5.204240848169634e-06, "loss": 0.0297, "step": 11998 }, { "epoch": 3.72, "learning_rate": 5.20384076815363e-06, "loss": 0.0288, "step": 11999 }, { "epoch": 3.72, "learning_rate": 5.203440688137628e-06, "loss": 0.1711, "step": 12000 }, { "epoch": 3.72, "learning_rate": 5.2030406081216245e-06, "loss": 0.1458, "step": 12001 }, { "epoch": 3.73, "learning_rate": 5.202640528105621e-06, "loss": 0.0291, "step": 12002 }, { "epoch": 3.73, "learning_rate": 5.202240448089618e-06, "loss": 0.051, "step": 12003 }, { "epoch": 3.73, "learning_rate": 5.2018403680736154e-06, "loss": 0.0286, "step": 12004 }, { "epoch": 3.73, "learning_rate": 5.201440288057612e-06, "loss": 0.0603, "step": 12005 }, { "epoch": 3.73, "learning_rate": 5.201040208041609e-06, "loss": 0.1304, "step": 12006 }, { "epoch": 3.73, "learning_rate": 5.2006401280256055e-06, "loss": 0.0732, "step": 12007 }, { "epoch": 3.73, "learning_rate": 5.200240048009603e-06, "loss": 0.0399, "step": 12008 }, { "epoch": 3.73, "learning_rate": 5.1998399679936e-06, "loss": 0.0479, "step": 12009 }, { "epoch": 3.73, "learning_rate": 5.199439887977596e-06, "loss": 0.1304, "step": 12010 }, { "epoch": 3.73, "learning_rate": 5.199039807961592e-06, "loss": 0.0182, "step": 12011 }, { "epoch": 3.73, "learning_rate": 5.198639727945589e-06, "loss": 0.0547, "step": 12012 }, { "epoch": 3.73, "learning_rate": 5.198239647929586e-06, "loss": 0.075, "step": 12013 }, { "epoch": 3.73, "learning_rate": 5.197839567913583e-06, "loss": 0.0555, "step": 12014 }, { "epoch": 3.73, "learning_rate": 5.19743948789758e-06, "loss": 0.066, "step": 12015 }, { "epoch": 3.73, "learning_rate": 5.1970394078815765e-06, "loss": 0.0506, "step": 12016 }, { "epoch": 3.73, "learning_rate": 5.196639327865573e-06, "loss": 0.1353, "step": 12017 }, { "epoch": 3.73, "learning_rate": 5.196239247849571e-06, "loss": 0.0576, "step": 12018 }, { "epoch": 3.73, "learning_rate": 5.195839167833567e-06, "loss": 0.0349, "step": 12019 }, { "epoch": 3.73, "learning_rate": 5.195439087817564e-06, "loss": 0.0798, "step": 12020 }, { "epoch": 3.73, "learning_rate": 5.195039007801561e-06, "loss": 0.0285, "step": 12021 }, { "epoch": 3.73, "learning_rate": 5.194638927785558e-06, "loss": 0.1374, "step": 12022 }, { "epoch": 3.73, "learning_rate": 5.194238847769554e-06, "loss": 0.0695, "step": 12023 }, { "epoch": 3.73, "learning_rate": 5.193838767753551e-06, "loss": 0.0321, "step": 12024 }, { "epoch": 3.73, "learning_rate": 5.1934386877375475e-06, "loss": 0.0896, "step": 12025 }, { "epoch": 3.73, "learning_rate": 5.193038607721544e-06, "loss": 0.0276, "step": 12026 }, { "epoch": 3.73, "learning_rate": 5.192638527705541e-06, "loss": 0.0474, "step": 12027 }, { "epoch": 3.73, "learning_rate": 5.1922384476895384e-06, "loss": 0.0307, "step": 12028 }, { "epoch": 3.73, "learning_rate": 5.191838367673535e-06, "loss": 0.0309, "step": 12029 }, { "epoch": 3.73, "learning_rate": 5.191438287657532e-06, "loss": 0.0513, "step": 12030 }, { "epoch": 3.73, "learning_rate": 5.1910382076415285e-06, "loss": 0.0813, "step": 12031 }, { "epoch": 3.73, "learning_rate": 5.190638127625526e-06, "loss": 0.1251, "step": 12032 }, { "epoch": 3.73, "learning_rate": 5.190238047609523e-06, "loss": 0.0196, "step": 12033 }, { "epoch": 3.74, "learning_rate": 5.189837967593519e-06, "loss": 0.077, "step": 12034 }, { "epoch": 3.74, "learning_rate": 5.189437887577516e-06, "loss": 0.0267, "step": 12035 }, { "epoch": 3.74, "learning_rate": 5.189037807561512e-06, "loss": 0.0156, "step": 12036 }, { "epoch": 3.74, "learning_rate": 5.1886377275455095e-06, "loss": 0.0323, "step": 12037 }, { "epoch": 3.74, "learning_rate": 5.188237647529506e-06, "loss": 0.0312, "step": 12038 }, { "epoch": 3.74, "learning_rate": 5.187837567513503e-06, "loss": 0.0325, "step": 12039 }, { "epoch": 3.74, "learning_rate": 5.1874374874974995e-06, "loss": 0.0624, "step": 12040 }, { "epoch": 3.74, "learning_rate": 5.187037407481497e-06, "loss": 0.1514, "step": 12041 }, { "epoch": 3.74, "learning_rate": 5.186637327465494e-06, "loss": 0.0228, "step": 12042 }, { "epoch": 3.74, "learning_rate": 5.1862372474494904e-06, "loss": 0.0282, "step": 12043 }, { "epoch": 3.74, "learning_rate": 5.185837167433487e-06, "loss": 0.0169, "step": 12044 }, { "epoch": 3.74, "learning_rate": 5.185437087417484e-06, "loss": 0.0623, "step": 12045 }, { "epoch": 3.74, "learning_rate": 5.185037007401481e-06, "loss": 0.0564, "step": 12046 }, { "epoch": 3.74, "learning_rate": 5.184636927385478e-06, "loss": 0.0445, "step": 12047 }, { "epoch": 3.74, "learning_rate": 5.184236847369475e-06, "loss": 0.037, "step": 12048 }, { "epoch": 3.74, "learning_rate": 5.1838367673534705e-06, "loss": 0.0594, "step": 12049 }, { "epoch": 3.74, "learning_rate": 5.183436687337467e-06, "loss": 0.0607, "step": 12050 }, { "epoch": 3.74, "learning_rate": 5.183036607321465e-06, "loss": 0.0443, "step": 12051 }, { "epoch": 3.74, "learning_rate": 5.1826365273054615e-06, "loss": 0.026, "step": 12052 }, { "epoch": 3.74, "learning_rate": 5.182236447289458e-06, "loss": 0.0525, "step": 12053 }, { "epoch": 3.74, "learning_rate": 5.181836367273455e-06, "loss": 0.0133, "step": 12054 }, { "epoch": 3.74, "learning_rate": 5.181436287257452e-06, "loss": 0.0474, "step": 12055 }, { "epoch": 3.74, "learning_rate": 5.181036207241449e-06, "loss": 0.0236, "step": 12056 }, { "epoch": 3.74, "learning_rate": 5.180636127225446e-06, "loss": 0.0309, "step": 12057 }, { "epoch": 3.74, "learning_rate": 5.180236047209442e-06, "loss": 0.0559, "step": 12058 }, { "epoch": 3.74, "learning_rate": 5.179835967193439e-06, "loss": 0.0299, "step": 12059 }, { "epoch": 3.74, "learning_rate": 5.179435887177437e-06, "loss": 0.091, "step": 12060 }, { "epoch": 3.74, "learning_rate": 5.1790358071614325e-06, "loss": 0.0351, "step": 12061 }, { "epoch": 3.74, "learning_rate": 5.178635727145429e-06, "loss": 0.1111, "step": 12062 }, { "epoch": 3.74, "learning_rate": 5.178235647129426e-06, "loss": 0.0364, "step": 12063 }, { "epoch": 3.74, "learning_rate": 5.1778355671134225e-06, "loss": 0.0193, "step": 12064 }, { "epoch": 3.74, "learning_rate": 5.17743548709742e-06, "loss": 0.0529, "step": 12065 }, { "epoch": 3.75, "learning_rate": 5.177035407081417e-06, "loss": 0.023, "step": 12066 }, { "epoch": 3.75, "learning_rate": 5.1766353270654134e-06, "loss": 0.0628, "step": 12067 }, { "epoch": 3.75, "learning_rate": 5.17623524704941e-06, "loss": 0.0255, "step": 12068 }, { "epoch": 3.75, "learning_rate": 5.175835167033408e-06, "loss": 0.0469, "step": 12069 }, { "epoch": 3.75, "learning_rate": 5.175435087017404e-06, "loss": 0.0425, "step": 12070 }, { "epoch": 3.75, "learning_rate": 5.175035007001401e-06, "loss": 0.0522, "step": 12071 }, { "epoch": 3.75, "learning_rate": 5.174634926985398e-06, "loss": 0.0348, "step": 12072 }, { "epoch": 3.75, "learning_rate": 5.174234846969394e-06, "loss": 0.1679, "step": 12073 }, { "epoch": 3.75, "learning_rate": 5.17383476695339e-06, "loss": 0.0406, "step": 12074 }, { "epoch": 3.75, "learning_rate": 5.173434686937388e-06, "loss": 0.0228, "step": 12075 }, { "epoch": 3.75, "learning_rate": 5.1730346069213845e-06, "loss": 0.0318, "step": 12076 }, { "epoch": 3.75, "learning_rate": 5.172634526905381e-06, "loss": 0.031, "step": 12077 }, { "epoch": 3.75, "learning_rate": 5.172234446889378e-06, "loss": 0.0427, "step": 12078 }, { "epoch": 3.75, "learning_rate": 5.171834366873375e-06, "loss": 0.0513, "step": 12079 }, { "epoch": 3.75, "learning_rate": 5.171434286857372e-06, "loss": 0.018, "step": 12080 }, { "epoch": 3.75, "learning_rate": 5.171034206841369e-06, "loss": 0.0194, "step": 12081 }, { "epoch": 3.75, "learning_rate": 5.1706341268253654e-06, "loss": 0.0533, "step": 12082 }, { "epoch": 3.75, "learning_rate": 5.170234046809363e-06, "loss": 0.0253, "step": 12083 }, { "epoch": 3.75, "learning_rate": 5.16983396679336e-06, "loss": 0.0405, "step": 12084 }, { "epoch": 3.75, "learning_rate": 5.169433886777356e-06, "loss": 0.0205, "step": 12085 }, { "epoch": 3.75, "learning_rate": 5.169033806761353e-06, "loss": 0.0776, "step": 12086 }, { "epoch": 3.75, "learning_rate": 5.168633726745349e-06, "loss": 0.0286, "step": 12087 }, { "epoch": 3.75, "learning_rate": 5.1682336467293455e-06, "loss": 0.0669, "step": 12088 }, { "epoch": 3.75, "learning_rate": 5.167833566713343e-06, "loss": 0.0546, "step": 12089 }, { "epoch": 3.75, "learning_rate": 5.16743348669734e-06, "loss": 0.0215, "step": 12090 }, { "epoch": 3.75, "learning_rate": 5.1670334066813365e-06, "loss": 0.0335, "step": 12091 }, { "epoch": 3.75, "learning_rate": 5.166633326665333e-06, "loss": 0.0308, "step": 12092 }, { "epoch": 3.75, "learning_rate": 5.166233246649331e-06, "loss": 0.0418, "step": 12093 }, { "epoch": 3.75, "learning_rate": 5.165833166633327e-06, "loss": 0.026, "step": 12094 }, { "epoch": 3.75, "learning_rate": 5.165433086617324e-06, "loss": 0.1156, "step": 12095 }, { "epoch": 3.75, "learning_rate": 5.165033006601321e-06, "loss": 0.04, "step": 12096 }, { "epoch": 3.75, "learning_rate": 5.164632926585318e-06, "loss": 0.0325, "step": 12097 }, { "epoch": 3.75, "learning_rate": 5.164232846569315e-06, "loss": 0.0668, "step": 12098 }, { "epoch": 3.76, "learning_rate": 5.163832766553311e-06, "loss": 0.0578, "step": 12099 }, { "epoch": 3.76, "learning_rate": 5.1634326865373075e-06, "loss": 0.0258, "step": 12100 }, { "epoch": 3.76, "learning_rate": 5.163032606521304e-06, "loss": 0.0685, "step": 12101 }, { "epoch": 3.76, "learning_rate": 5.162632526505301e-06, "loss": 0.0879, "step": 12102 }, { "epoch": 3.76, "learning_rate": 5.162232446489298e-06, "loss": 0.0485, "step": 12103 }, { "epoch": 3.76, "learning_rate": 5.161832366473295e-06, "loss": 0.0322, "step": 12104 }, { "epoch": 3.76, "learning_rate": 5.161432286457292e-06, "loss": 0.0264, "step": 12105 }, { "epoch": 3.76, "learning_rate": 5.1610322064412884e-06, "loss": 0.0727, "step": 12106 }, { "epoch": 3.76, "learning_rate": 5.160632126425286e-06, "loss": 0.0274, "step": 12107 }, { "epoch": 3.76, "learning_rate": 5.160232046409283e-06, "loss": 0.045, "step": 12108 }, { "epoch": 3.76, "learning_rate": 5.159831966393279e-06, "loss": 0.0365, "step": 12109 }, { "epoch": 3.76, "learning_rate": 5.159431886377276e-06, "loss": 0.0398, "step": 12110 }, { "epoch": 3.76, "learning_rate": 5.1590318063612736e-06, "loss": 0.149, "step": 12111 }, { "epoch": 3.76, "learning_rate": 5.158631726345269e-06, "loss": 0.0329, "step": 12112 }, { "epoch": 3.76, "learning_rate": 5.158231646329266e-06, "loss": 0.0229, "step": 12113 }, { "epoch": 3.76, "learning_rate": 5.157831566313263e-06, "loss": 0.028, "step": 12114 }, { "epoch": 3.76, "learning_rate": 5.1574314862972595e-06, "loss": 0.0164, "step": 12115 }, { "epoch": 3.76, "learning_rate": 5.157031406281256e-06, "loss": 0.1248, "step": 12116 }, { "epoch": 3.76, "learning_rate": 5.156631326265254e-06, "loss": 0.0497, "step": 12117 }, { "epoch": 3.76, "learning_rate": 5.15623124624925e-06, "loss": 0.0589, "step": 12118 }, { "epoch": 3.76, "learning_rate": 5.155831166233247e-06, "loss": 0.0325, "step": 12119 }, { "epoch": 3.76, "learning_rate": 5.155431086217244e-06, "loss": 0.0429, "step": 12120 }, { "epoch": 3.76, "learning_rate": 5.155031006201241e-06, "loss": 0.0816, "step": 12121 }, { "epoch": 3.76, "learning_rate": 5.154630926185238e-06, "loss": 0.0357, "step": 12122 }, { "epoch": 3.76, "learning_rate": 5.154230846169235e-06, "loss": 0.0413, "step": 12123 }, { "epoch": 3.76, "learning_rate": 5.1538307661532305e-06, "loss": 0.0514, "step": 12124 }, { "epoch": 3.76, "learning_rate": 5.153430686137227e-06, "loss": 0.0498, "step": 12125 }, { "epoch": 3.76, "learning_rate": 5.153030606121225e-06, "loss": 0.0289, "step": 12126 }, { "epoch": 3.76, "learning_rate": 5.152630526105221e-06, "loss": 0.0366, "step": 12127 }, { "epoch": 3.76, "learning_rate": 5.152230446089218e-06, "loss": 0.0411, "step": 12128 }, { "epoch": 3.76, "learning_rate": 5.151830366073215e-06, "loss": 0.046, "step": 12129 }, { "epoch": 3.76, "learning_rate": 5.1514302860572115e-06, "loss": 0.0384, "step": 12130 }, { "epoch": 3.77, "learning_rate": 5.151030206041209e-06, "loss": 0.0489, "step": 12131 }, { "epoch": 3.77, "learning_rate": 5.150630126025206e-06, "loss": 0.0254, "step": 12132 }, { "epoch": 3.77, "learning_rate": 5.150230046009202e-06, "loss": 0.0325, "step": 12133 }, { "epoch": 3.77, "learning_rate": 5.149829965993199e-06, "loss": 0.0349, "step": 12134 }, { "epoch": 3.77, "learning_rate": 5.149429885977197e-06, "loss": 0.057, "step": 12135 }, { "epoch": 3.77, "learning_rate": 5.149029805961193e-06, "loss": 0.0706, "step": 12136 }, { "epoch": 3.77, "learning_rate": 5.148629725945189e-06, "loss": 0.043, "step": 12137 }, { "epoch": 3.77, "learning_rate": 5.148229645929186e-06, "loss": 0.0234, "step": 12138 }, { "epoch": 3.77, "learning_rate": 5.1478295659131825e-06, "loss": 0.0271, "step": 12139 }, { "epoch": 3.77, "learning_rate": 5.14742948589718e-06, "loss": 0.0394, "step": 12140 }, { "epoch": 3.77, "learning_rate": 5.147029405881177e-06, "loss": 0.0423, "step": 12141 }, { "epoch": 3.77, "learning_rate": 5.146629325865173e-06, "loss": 0.033, "step": 12142 }, { "epoch": 3.77, "learning_rate": 5.14622924584917e-06, "loss": 0.0408, "step": 12143 }, { "epoch": 3.77, "learning_rate": 5.145829165833167e-06, "loss": 0.0849, "step": 12144 }, { "epoch": 3.77, "learning_rate": 5.145429085817164e-06, "loss": 0.0174, "step": 12145 }, { "epoch": 3.77, "learning_rate": 5.145029005801161e-06, "loss": 0.0395, "step": 12146 }, { "epoch": 3.77, "learning_rate": 5.144628925785158e-06, "loss": 0.0132, "step": 12147 }, { "epoch": 3.77, "learning_rate": 5.144228845769154e-06, "loss": 0.029, "step": 12148 }, { "epoch": 3.77, "learning_rate": 5.143828765753152e-06, "loss": 0.0196, "step": 12149 }, { "epoch": 3.77, "learning_rate": 5.143428685737148e-06, "loss": 0.0783, "step": 12150 }, { "epoch": 3.77, "learning_rate": 5.143028605721144e-06, "loss": 0.0363, "step": 12151 }, { "epoch": 3.77, "learning_rate": 5.142628525705141e-06, "loss": 0.045, "step": 12152 }, { "epoch": 3.77, "learning_rate": 5.142228445689138e-06, "loss": 0.0144, "step": 12153 }, { "epoch": 3.77, "learning_rate": 5.141828365673135e-06, "loss": 0.074, "step": 12154 }, { "epoch": 3.77, "learning_rate": 5.141428285657132e-06, "loss": 0.0147, "step": 12155 }, { "epoch": 3.77, "learning_rate": 5.141028205641129e-06, "loss": 0.0164, "step": 12156 }, { "epoch": 3.77, "learning_rate": 5.140628125625125e-06, "loss": 0.0442, "step": 12157 }, { "epoch": 3.77, "learning_rate": 5.140228045609122e-06, "loss": 0.0348, "step": 12158 }, { "epoch": 3.77, "learning_rate": 5.13982796559312e-06, "loss": 0.0354, "step": 12159 }, { "epoch": 3.77, "learning_rate": 5.139427885577116e-06, "loss": 0.0347, "step": 12160 }, { "epoch": 3.77, "learning_rate": 5.139027805561113e-06, "loss": 0.1037, "step": 12161 }, { "epoch": 3.77, "learning_rate": 5.138627725545109e-06, "loss": 0.021, "step": 12162 }, { "epoch": 3.78, "learning_rate": 5.1382276455291055e-06, "loss": 0.0507, "step": 12163 }, { "epoch": 3.78, "learning_rate": 5.137827565513103e-06, "loss": 0.0573, "step": 12164 }, { "epoch": 3.78, "learning_rate": 5.1374274854971e-06, "loss": 0.0343, "step": 12165 }, { "epoch": 3.78, "learning_rate": 5.137027405481096e-06, "loss": 0.0546, "step": 12166 }, { "epoch": 3.78, "learning_rate": 5.136627325465093e-06, "loss": 0.0241, "step": 12167 }, { "epoch": 3.78, "learning_rate": 5.136227245449091e-06, "loss": 0.0489, "step": 12168 }, { "epoch": 3.78, "learning_rate": 5.135827165433087e-06, "loss": 0.0594, "step": 12169 }, { "epoch": 3.78, "learning_rate": 5.135427085417084e-06, "loss": 0.0386, "step": 12170 }, { "epoch": 3.78, "learning_rate": 5.135027005401081e-06, "loss": 0.0363, "step": 12171 }, { "epoch": 3.78, "learning_rate": 5.134626925385078e-06, "loss": 0.1303, "step": 12172 }, { "epoch": 3.78, "learning_rate": 5.134226845369075e-06, "loss": 0.0244, "step": 12173 }, { "epoch": 3.78, "learning_rate": 5.133826765353072e-06, "loss": 0.049, "step": 12174 }, { "epoch": 3.78, "learning_rate": 5.133426685337067e-06, "loss": 0.0435, "step": 12175 }, { "epoch": 3.78, "learning_rate": 5.133026605321064e-06, "loss": 0.0506, "step": 12176 }, { "epoch": 3.78, "learning_rate": 5.132626525305061e-06, "loss": 0.0529, "step": 12177 }, { "epoch": 3.78, "learning_rate": 5.132226445289058e-06, "loss": 0.0772, "step": 12178 }, { "epoch": 3.78, "learning_rate": 5.131826365273055e-06, "loss": 0.0466, "step": 12179 }, { "epoch": 3.78, "learning_rate": 5.131426285257052e-06, "loss": 0.0577, "step": 12180 }, { "epoch": 3.78, "learning_rate": 5.131026205241048e-06, "loss": 0.0484, "step": 12181 }, { "epoch": 3.78, "learning_rate": 5.130626125225046e-06, "loss": 0.0467, "step": 12182 }, { "epoch": 3.78, "learning_rate": 5.130226045209043e-06, "loss": 0.0567, "step": 12183 }, { "epoch": 3.78, "learning_rate": 5.129825965193039e-06, "loss": 0.0285, "step": 12184 }, { "epoch": 3.78, "learning_rate": 5.129425885177036e-06, "loss": 0.0159, "step": 12185 }, { "epoch": 3.78, "learning_rate": 5.1290258051610335e-06, "loss": 0.0339, "step": 12186 }, { "epoch": 3.78, "learning_rate": 5.1286257251450285e-06, "loss": 0.0564, "step": 12187 }, { "epoch": 3.78, "learning_rate": 5.128225645129026e-06, "loss": 0.0683, "step": 12188 }, { "epoch": 3.78, "learning_rate": 5.127825565113023e-06, "loss": 0.1393, "step": 12189 }, { "epoch": 3.78, "learning_rate": 5.127425485097019e-06, "loss": 0.0324, "step": 12190 }, { "epoch": 3.78, "learning_rate": 5.127025405081016e-06, "loss": 0.0431, "step": 12191 }, { "epoch": 3.78, "learning_rate": 5.126625325065014e-06, "loss": 0.0225, "step": 12192 }, { "epoch": 3.78, "learning_rate": 5.12622524504901e-06, "loss": 0.0282, "step": 12193 }, { "epoch": 3.78, "learning_rate": 5.125825165033007e-06, "loss": 0.0307, "step": 12194 }, { "epoch": 3.79, "learning_rate": 5.125425085017004e-06, "loss": 0.0435, "step": 12195 }, { "epoch": 3.79, "learning_rate": 5.125025005001001e-06, "loss": 0.038, "step": 12196 }, { "epoch": 3.79, "learning_rate": 5.124624924984998e-06, "loss": 0.1635, "step": 12197 }, { "epoch": 3.79, "learning_rate": 5.124224844968995e-06, "loss": 0.1107, "step": 12198 }, { "epoch": 3.79, "learning_rate": 5.123824764952991e-06, "loss": 0.1021, "step": 12199 }, { "epoch": 3.79, "learning_rate": 5.123424684936987e-06, "loss": 0.097, "step": 12200 }, { "epoch": 3.79, "learning_rate": 5.123024604920984e-06, "loss": 0.0539, "step": 12201 }, { "epoch": 3.79, "learning_rate": 5.122624524904981e-06, "loss": 0.0419, "step": 12202 }, { "epoch": 3.79, "learning_rate": 5.122224444888978e-06, "loss": 0.0919, "step": 12203 }, { "epoch": 3.79, "learning_rate": 5.121824364872975e-06, "loss": 0.02, "step": 12204 }, { "epoch": 3.79, "learning_rate": 5.121424284856971e-06, "loss": 0.0195, "step": 12205 }, { "epoch": 3.79, "learning_rate": 5.121024204840969e-06, "loss": 0.0301, "step": 12206 }, { "epoch": 3.79, "learning_rate": 5.120624124824966e-06, "loss": 0.025, "step": 12207 }, { "epoch": 3.79, "learning_rate": 5.120224044808962e-06, "loss": 0.0402, "step": 12208 }, { "epoch": 3.79, "learning_rate": 5.119823964792959e-06, "loss": 0.0438, "step": 12209 }, { "epoch": 3.79, "learning_rate": 5.1194238847769565e-06, "loss": 0.0652, "step": 12210 }, { "epoch": 3.79, "learning_rate": 5.119023804760953e-06, "loss": 0.0391, "step": 12211 }, { "epoch": 3.79, "learning_rate": 5.11862372474495e-06, "loss": 0.0214, "step": 12212 }, { "epoch": 3.79, "learning_rate": 5.118223644728946e-06, "loss": 0.0192, "step": 12213 }, { "epoch": 3.79, "learning_rate": 5.117823564712942e-06, "loss": 0.062, "step": 12214 }, { "epoch": 3.79, "learning_rate": 5.117423484696939e-06, "loss": 0.0373, "step": 12215 }, { "epoch": 3.79, "learning_rate": 5.117023404680937e-06, "loss": 0.0481, "step": 12216 }, { "epoch": 3.79, "learning_rate": 5.116623324664933e-06, "loss": 0.03, "step": 12217 }, { "epoch": 3.79, "learning_rate": 5.11622324464893e-06, "loss": 0.0575, "step": 12218 }, { "epoch": 3.79, "learning_rate": 5.115823164632927e-06, "loss": 0.1376, "step": 12219 }, { "epoch": 3.79, "learning_rate": 5.115423084616924e-06, "loss": 0.0294, "step": 12220 }, { "epoch": 3.79, "learning_rate": 5.115023004600921e-06, "loss": 0.0416, "step": 12221 }, { "epoch": 3.79, "learning_rate": 5.114622924584918e-06, "loss": 0.0979, "step": 12222 }, { "epoch": 3.79, "learning_rate": 5.114222844568914e-06, "loss": 0.0384, "step": 12223 }, { "epoch": 3.79, "learning_rate": 5.113822764552912e-06, "loss": 0.0422, "step": 12224 }, { "epoch": 3.79, "learning_rate": 5.113422684536908e-06, "loss": 0.0438, "step": 12225 }, { "epoch": 3.79, "learning_rate": 5.113022604520904e-06, "loss": 0.0474, "step": 12226 }, { "epoch": 3.79, "learning_rate": 5.112622524504901e-06, "loss": 0.0242, "step": 12227 }, { "epoch": 3.8, "learning_rate": 5.112222444488898e-06, "loss": 0.0296, "step": 12228 }, { "epoch": 3.8, "learning_rate": 5.111822364472894e-06, "loss": 0.0408, "step": 12229 }, { "epoch": 3.8, "learning_rate": 5.111422284456892e-06, "loss": 0.0251, "step": 12230 }, { "epoch": 3.8, "learning_rate": 5.111022204440889e-06, "loss": 0.0357, "step": 12231 }, { "epoch": 3.8, "learning_rate": 5.110622124424885e-06, "loss": 0.0375, "step": 12232 }, { "epoch": 3.8, "learning_rate": 5.110222044408882e-06, "loss": 0.0864, "step": 12233 }, { "epoch": 3.8, "learning_rate": 5.1098219643928795e-06, "loss": 0.03, "step": 12234 }, { "epoch": 3.8, "learning_rate": 5.109421884376876e-06, "loss": 0.0494, "step": 12235 }, { "epoch": 3.8, "learning_rate": 5.109021804360873e-06, "loss": 0.0331, "step": 12236 }, { "epoch": 3.8, "learning_rate": 5.10862172434487e-06, "loss": 0.0213, "step": 12237 }, { "epoch": 3.8, "learning_rate": 5.1082216443288654e-06, "loss": 0.0207, "step": 12238 }, { "epoch": 3.8, "learning_rate": 5.107821564312863e-06, "loss": 0.0229, "step": 12239 }, { "epoch": 3.8, "learning_rate": 5.10742148429686e-06, "loss": 0.0443, "step": 12240 }, { "epoch": 3.8, "learning_rate": 5.107021404280856e-06, "loss": 0.0447, "step": 12241 }, { "epoch": 3.8, "learning_rate": 5.106621324264853e-06, "loss": 0.028, "step": 12242 }, { "epoch": 3.8, "learning_rate": 5.10622124424885e-06, "loss": 0.0864, "step": 12243 }, { "epoch": 3.8, "learning_rate": 5.105821164232847e-06, "loss": 0.1298, "step": 12244 }, { "epoch": 3.8, "learning_rate": 5.105421084216844e-06, "loss": 0.0145, "step": 12245 }, { "epoch": 3.8, "learning_rate": 5.105021004200841e-06, "loss": 0.0571, "step": 12246 }, { "epoch": 3.8, "learning_rate": 5.104620924184837e-06, "loss": 0.0254, "step": 12247 }, { "epoch": 3.8, "learning_rate": 5.104220844168835e-06, "loss": 0.0328, "step": 12248 }, { "epoch": 3.8, "learning_rate": 5.1038207641528315e-06, "loss": 0.0347, "step": 12249 }, { "epoch": 3.8, "learning_rate": 5.103420684136827e-06, "loss": 0.0409, "step": 12250 }, { "epoch": 3.8, "learning_rate": 5.103020604120824e-06, "loss": 0.0328, "step": 12251 }, { "epoch": 3.8, "learning_rate": 5.102620524104821e-06, "loss": 0.0582, "step": 12252 }, { "epoch": 3.8, "learning_rate": 5.102220444088818e-06, "loss": 0.0321, "step": 12253 }, { "epoch": 3.8, "learning_rate": 5.101820364072815e-06, "loss": 0.0563, "step": 12254 }, { "epoch": 3.8, "learning_rate": 5.101420284056812e-06, "loss": 0.0463, "step": 12255 }, { "epoch": 3.8, "learning_rate": 5.101020204040808e-06, "loss": 0.0381, "step": 12256 }, { "epoch": 3.8, "learning_rate": 5.100620124024806e-06, "loss": 0.0384, "step": 12257 }, { "epoch": 3.8, "learning_rate": 5.1002200440088025e-06, "loss": 0.0502, "step": 12258 }, { "epoch": 3.8, "learning_rate": 5.099819963992799e-06, "loss": 0.0829, "step": 12259 }, { "epoch": 3.81, "learning_rate": 5.099419883976796e-06, "loss": 0.05, "step": 12260 }, { "epoch": 3.81, "learning_rate": 5.099019803960793e-06, "loss": 0.0935, "step": 12261 }, { "epoch": 3.81, "learning_rate": 5.09861972394479e-06, "loss": 0.0251, "step": 12262 }, { "epoch": 3.81, "learning_rate": 5.098219643928786e-06, "loss": 0.0621, "step": 12263 }, { "epoch": 3.81, "learning_rate": 5.097819563912783e-06, "loss": 0.0346, "step": 12264 }, { "epoch": 3.81, "learning_rate": 5.097419483896779e-06, "loss": 0.0322, "step": 12265 }, { "epoch": 3.81, "learning_rate": 5.097019403880776e-06, "loss": 0.0228, "step": 12266 }, { "epoch": 3.81, "learning_rate": 5.0966193238647736e-06, "loss": 0.073, "step": 12267 }, { "epoch": 3.81, "learning_rate": 5.09621924384877e-06, "loss": 0.0432, "step": 12268 }, { "epoch": 3.81, "learning_rate": 5.095819163832767e-06, "loss": 0.0399, "step": 12269 }, { "epoch": 3.81, "learning_rate": 5.095419083816764e-06, "loss": 0.0955, "step": 12270 }, { "epoch": 3.81, "learning_rate": 5.095019003800761e-06, "loss": 0.0246, "step": 12271 }, { "epoch": 3.81, "learning_rate": 5.094618923784758e-06, "loss": 0.0162, "step": 12272 }, { "epoch": 3.81, "learning_rate": 5.0942188437687545e-06, "loss": 0.0804, "step": 12273 }, { "epoch": 3.81, "learning_rate": 5.093818763752751e-06, "loss": 0.0993, "step": 12274 }, { "epoch": 3.81, "learning_rate": 5.093418683736748e-06, "loss": 0.045, "step": 12275 }, { "epoch": 3.81, "learning_rate": 5.093018603720744e-06, "loss": 0.0336, "step": 12276 }, { "epoch": 3.81, "learning_rate": 5.092618523704741e-06, "loss": 0.0121, "step": 12277 }, { "epoch": 3.81, "learning_rate": 5.092218443688738e-06, "loss": 0.0422, "step": 12278 }, { "epoch": 3.81, "learning_rate": 5.091818363672735e-06, "loss": 0.032, "step": 12279 }, { "epoch": 3.81, "learning_rate": 5.091418283656731e-06, "loss": 0.0266, "step": 12280 }, { "epoch": 3.81, "learning_rate": 5.091018203640729e-06, "loss": 0.0392, "step": 12281 }, { "epoch": 3.81, "learning_rate": 5.0906181236247256e-06, "loss": 0.0281, "step": 12282 }, { "epoch": 3.81, "learning_rate": 5.090218043608722e-06, "loss": 0.0466, "step": 12283 }, { "epoch": 3.81, "learning_rate": 5.089817963592719e-06, "loss": 0.0238, "step": 12284 }, { "epoch": 3.81, "learning_rate": 5.0894178835767165e-06, "loss": 0.0301, "step": 12285 }, { "epoch": 3.81, "learning_rate": 5.089017803560713e-06, "loss": 0.0176, "step": 12286 }, { "epoch": 3.81, "learning_rate": 5.08861772354471e-06, "loss": 0.0487, "step": 12287 }, { "epoch": 3.81, "learning_rate": 5.088217643528706e-06, "loss": 0.034, "step": 12288 }, { "epoch": 3.81, "learning_rate": 5.087817563512702e-06, "loss": 0.0453, "step": 12289 }, { "epoch": 3.81, "learning_rate": 5.087417483496699e-06, "loss": 0.0391, "step": 12290 }, { "epoch": 3.81, "learning_rate": 5.087017403480697e-06, "loss": 0.0177, "step": 12291 }, { "epoch": 3.82, "learning_rate": 5.086617323464693e-06, "loss": 0.0439, "step": 12292 }, { "epoch": 3.82, "learning_rate": 5.08621724344869e-06, "loss": 0.0364, "step": 12293 }, { "epoch": 3.82, "learning_rate": 5.085817163432687e-06, "loss": 0.0666, "step": 12294 }, { "epoch": 3.82, "learning_rate": 5.085417083416684e-06, "loss": 0.0146, "step": 12295 }, { "epoch": 3.82, "learning_rate": 5.085017003400681e-06, "loss": 0.054, "step": 12296 }, { "epoch": 3.82, "learning_rate": 5.0846169233846775e-06, "loss": 0.0198, "step": 12297 }, { "epoch": 3.82, "learning_rate": 5.084216843368674e-06, "loss": 0.0336, "step": 12298 }, { "epoch": 3.82, "learning_rate": 5.083816763352672e-06, "loss": 0.0516, "step": 12299 }, { "epoch": 3.82, "learning_rate": 5.0834166833366685e-06, "loss": 0.0502, "step": 12300 }, { "epoch": 3.82, "learning_rate": 5.083016603320664e-06, "loss": 0.045, "step": 12301 }, { "epoch": 3.82, "learning_rate": 5.082616523304661e-06, "loss": 0.0438, "step": 12302 }, { "epoch": 3.82, "learning_rate": 5.082216443288658e-06, "loss": 0.0652, "step": 12303 }, { "epoch": 3.82, "learning_rate": 5.081816363272654e-06, "loss": 0.0556, "step": 12304 }, { "epoch": 3.82, "learning_rate": 5.081416283256652e-06, "loss": 0.0413, "step": 12305 }, { "epoch": 3.82, "learning_rate": 5.0810162032406486e-06, "loss": 0.06, "step": 12306 }, { "epoch": 3.82, "learning_rate": 5.080616123224645e-06, "loss": 0.0351, "step": 12307 }, { "epoch": 3.82, "learning_rate": 5.080216043208642e-06, "loss": 0.0596, "step": 12308 }, { "epoch": 3.82, "learning_rate": 5.0798159631926395e-06, "loss": 0.0139, "step": 12309 }, { "epoch": 3.82, "learning_rate": 5.079415883176636e-06, "loss": 0.0252, "step": 12310 }, { "epoch": 3.82, "learning_rate": 5.079015803160633e-06, "loss": 0.1953, "step": 12311 }, { "epoch": 3.82, "learning_rate": 5.0786157231446295e-06, "loss": 0.0705, "step": 12312 }, { "epoch": 3.82, "learning_rate": 5.078215643128627e-06, "loss": 0.0217, "step": 12313 }, { "epoch": 3.82, "learning_rate": 5.077815563112622e-06, "loss": 0.0292, "step": 12314 }, { "epoch": 3.82, "learning_rate": 5.07741548309662e-06, "loss": 0.0414, "step": 12315 }, { "epoch": 3.82, "learning_rate": 5.077015403080616e-06, "loss": 0.0464, "step": 12316 }, { "epoch": 3.82, "learning_rate": 5.076615323064613e-06, "loss": 0.0579, "step": 12317 }, { "epoch": 3.82, "learning_rate": 5.07621524304861e-06, "loss": 0.0284, "step": 12318 }, { "epoch": 3.82, "learning_rate": 5.075815163032607e-06, "loss": 0.0339, "step": 12319 }, { "epoch": 3.82, "learning_rate": 5.075415083016604e-06, "loss": 0.0272, "step": 12320 }, { "epoch": 3.82, "learning_rate": 5.0750150030006006e-06, "loss": 0.0154, "step": 12321 }, { "epoch": 3.82, "learning_rate": 5.074614922984597e-06, "loss": 0.0275, "step": 12322 }, { "epoch": 3.82, "learning_rate": 5.074214842968595e-06, "loss": 0.0533, "step": 12323 }, { "epoch": 3.83, "learning_rate": 5.0738147629525915e-06, "loss": 0.028, "step": 12324 }, { "epoch": 3.83, "learning_rate": 5.073414682936588e-06, "loss": 0.0208, "step": 12325 }, { "epoch": 3.83, "learning_rate": 5.073014602920584e-06, "loss": 0.0411, "step": 12326 }, { "epoch": 3.83, "learning_rate": 5.072614522904581e-06, "loss": 0.0547, "step": 12327 }, { "epoch": 3.83, "learning_rate": 5.072214442888577e-06, "loss": 0.035, "step": 12328 }, { "epoch": 3.83, "learning_rate": 5.071814362872575e-06, "loss": 0.0618, "step": 12329 }, { "epoch": 3.83, "learning_rate": 5.071414282856572e-06, "loss": 0.0497, "step": 12330 }, { "epoch": 3.83, "learning_rate": 5.071014202840568e-06, "loss": 0.0292, "step": 12331 }, { "epoch": 3.83, "learning_rate": 5.070614122824565e-06, "loss": 0.069, "step": 12332 }, { "epoch": 3.83, "learning_rate": 5.0702140428085625e-06, "loss": 0.0293, "step": 12333 }, { "epoch": 3.83, "learning_rate": 5.069813962792559e-06, "loss": 0.0398, "step": 12334 }, { "epoch": 3.83, "learning_rate": 5.069413882776556e-06, "loss": 0.1069, "step": 12335 }, { "epoch": 3.83, "learning_rate": 5.0690138027605525e-06, "loss": 0.0417, "step": 12336 }, { "epoch": 3.83, "learning_rate": 5.06861372274455e-06, "loss": 0.0271, "step": 12337 }, { "epoch": 3.83, "learning_rate": 5.068213642728547e-06, "loss": 0.0342, "step": 12338 }, { "epoch": 3.83, "learning_rate": 5.067813562712543e-06, "loss": 0.051, "step": 12339 }, { "epoch": 3.83, "learning_rate": 5.067413482696539e-06, "loss": 0.0417, "step": 12340 }, { "epoch": 3.83, "learning_rate": 5.067013402680536e-06, "loss": 0.0512, "step": 12341 }, { "epoch": 3.83, "learning_rate": 5.0666133226645335e-06, "loss": 0.039, "step": 12342 }, { "epoch": 3.83, "learning_rate": 5.06621324264853e-06, "loss": 0.0744, "step": 12343 }, { "epoch": 3.83, "learning_rate": 5.065813162632527e-06, "loss": 0.052, "step": 12344 }, { "epoch": 3.83, "learning_rate": 5.0654130826165236e-06, "loss": 0.0502, "step": 12345 }, { "epoch": 3.83, "learning_rate": 5.06501300260052e-06, "loss": 0.0437, "step": 12346 }, { "epoch": 3.83, "learning_rate": 5.064612922584518e-06, "loss": 0.0302, "step": 12347 }, { "epoch": 3.83, "learning_rate": 5.0642128425685145e-06, "loss": 0.0122, "step": 12348 }, { "epoch": 3.83, "learning_rate": 5.063812762552511e-06, "loss": 0.1008, "step": 12349 }, { "epoch": 3.83, "learning_rate": 5.063412682536508e-06, "loss": 0.0405, "step": 12350 }, { "epoch": 3.83, "learning_rate": 5.063012602520504e-06, "loss": 0.0435, "step": 12351 }, { "epoch": 3.83, "learning_rate": 5.062612522504501e-06, "loss": 0.0261, "step": 12352 }, { "epoch": 3.83, "learning_rate": 5.062212442488498e-06, "loss": 0.0503, "step": 12353 }, { "epoch": 3.83, "learning_rate": 5.061812362472495e-06, "loss": 0.0114, "step": 12354 }, { "epoch": 3.83, "learning_rate": 5.061412282456491e-06, "loss": 0.048, "step": 12355 }, { "epoch": 3.84, "learning_rate": 5.061012202440489e-06, "loss": 0.0271, "step": 12356 }, { "epoch": 3.84, "learning_rate": 5.0606121224244855e-06, "loss": 0.0675, "step": 12357 }, { "epoch": 3.84, "learning_rate": 5.060212042408482e-06, "loss": 0.0435, "step": 12358 }, { "epoch": 3.84, "learning_rate": 5.059811962392479e-06, "loss": 0.029, "step": 12359 }, { "epoch": 3.84, "learning_rate": 5.0594118823764756e-06, "loss": 0.0569, "step": 12360 }, { "epoch": 3.84, "learning_rate": 5.059011802360473e-06, "loss": 0.0347, "step": 12361 }, { "epoch": 3.84, "learning_rate": 5.05861172234447e-06, "loss": 0.0514, "step": 12362 }, { "epoch": 3.84, "learning_rate": 5.0582116423284665e-06, "loss": 0.0383, "step": 12363 }, { "epoch": 3.84, "learning_rate": 5.057811562312462e-06, "loss": 0.035, "step": 12364 }, { "epoch": 3.84, "learning_rate": 5.057411482296459e-06, "loss": 0.0148, "step": 12365 }, { "epoch": 3.84, "learning_rate": 5.0570114022804565e-06, "loss": 0.0501, "step": 12366 }, { "epoch": 3.84, "learning_rate": 5.056611322264453e-06, "loss": 0.0884, "step": 12367 }, { "epoch": 3.84, "learning_rate": 5.05621124224845e-06, "loss": 0.0222, "step": 12368 }, { "epoch": 3.84, "learning_rate": 5.055811162232447e-06, "loss": 0.1436, "step": 12369 }, { "epoch": 3.84, "learning_rate": 5.055411082216444e-06, "loss": 0.0892, "step": 12370 }, { "epoch": 3.84, "learning_rate": 5.055011002200441e-06, "loss": 0.0338, "step": 12371 }, { "epoch": 3.84, "learning_rate": 5.0546109221844375e-06, "loss": 0.0509, "step": 12372 }, { "epoch": 3.84, "learning_rate": 5.054210842168434e-06, "loss": 0.0097, "step": 12373 }, { "epoch": 3.84, "learning_rate": 5.053810762152431e-06, "loss": 0.0374, "step": 12374 }, { "epoch": 3.84, "learning_rate": 5.053410682136428e-06, "loss": 0.1213, "step": 12375 }, { "epoch": 3.84, "learning_rate": 5.053010602120425e-06, "loss": 0.0364, "step": 12376 }, { "epoch": 3.84, "learning_rate": 5.052610522104421e-06, "loss": 0.1026, "step": 12377 }, { "epoch": 3.84, "learning_rate": 5.052210442088418e-06, "loss": 0.034, "step": 12378 }, { "epoch": 3.84, "learning_rate": 5.051810362072414e-06, "loss": 0.1154, "step": 12379 }, { "epoch": 3.84, "learning_rate": 5.051410282056412e-06, "loss": 0.0614, "step": 12380 }, { "epoch": 3.84, "learning_rate": 5.0510102020404085e-06, "loss": 0.0535, "step": 12381 }, { "epoch": 3.84, "learning_rate": 5.050610122024405e-06, "loss": 0.0383, "step": 12382 }, { "epoch": 3.84, "learning_rate": 5.050210042008402e-06, "loss": 0.0333, "step": 12383 }, { "epoch": 3.84, "learning_rate": 5.049809961992399e-06, "loss": 0.0625, "step": 12384 }, { "epoch": 3.84, "learning_rate": 5.049409881976396e-06, "loss": 0.0374, "step": 12385 }, { "epoch": 3.84, "learning_rate": 5.049009801960393e-06, "loss": 0.065, "step": 12386 }, { "epoch": 3.84, "learning_rate": 5.0486097219443895e-06, "loss": 0.0659, "step": 12387 }, { "epoch": 3.84, "learning_rate": 5.048209641928387e-06, "loss": 0.0375, "step": 12388 }, { "epoch": 3.85, "learning_rate": 5.047809561912382e-06, "loss": 0.0747, "step": 12389 }, { "epoch": 3.85, "learning_rate": 5.0474094818963795e-06, "loss": 0.0422, "step": 12390 }, { "epoch": 3.85, "learning_rate": 5.047009401880376e-06, "loss": 0.0397, "step": 12391 }, { "epoch": 3.85, "learning_rate": 5.046609321864373e-06, "loss": 0.0251, "step": 12392 }, { "epoch": 3.85, "learning_rate": 5.04620924184837e-06, "loss": 0.0183, "step": 12393 }, { "epoch": 3.85, "learning_rate": 5.045809161832367e-06, "loss": 0.0289, "step": 12394 }, { "epoch": 3.85, "learning_rate": 5.045409081816364e-06, "loss": 0.0201, "step": 12395 }, { "epoch": 3.85, "learning_rate": 5.0450090018003605e-06, "loss": 0.0699, "step": 12396 }, { "epoch": 3.85, "learning_rate": 5.044608921784357e-06, "loss": 0.0293, "step": 12397 }, { "epoch": 3.85, "learning_rate": 5.044208841768355e-06, "loss": 0.1013, "step": 12398 }, { "epoch": 3.85, "learning_rate": 5.043808761752351e-06, "loss": 0.0363, "step": 12399 }, { "epoch": 3.85, "learning_rate": 5.043408681736348e-06, "loss": 0.0129, "step": 12400 }, { "epoch": 3.85, "learning_rate": 5.043008601720345e-06, "loss": 0.0364, "step": 12401 }, { "epoch": 3.85, "learning_rate": 5.042608521704341e-06, "loss": 0.0231, "step": 12402 }, { "epoch": 3.85, "learning_rate": 5.042208441688337e-06, "loss": 0.0468, "step": 12403 }, { "epoch": 3.85, "learning_rate": 5.041808361672335e-06, "loss": 0.0143, "step": 12404 }, { "epoch": 3.85, "learning_rate": 5.0414082816563315e-06, "loss": 0.0804, "step": 12405 }, { "epoch": 3.85, "learning_rate": 5.041008201640328e-06, "loss": 0.065, "step": 12406 }, { "epoch": 3.85, "learning_rate": 5.040608121624325e-06, "loss": 0.0369, "step": 12407 }, { "epoch": 3.85, "learning_rate": 5.0402080416083224e-06, "loss": 0.0553, "step": 12408 }, { "epoch": 3.85, "learning_rate": 5.039807961592319e-06, "loss": 0.0534, "step": 12409 }, { "epoch": 3.85, "learning_rate": 5.039407881576316e-06, "loss": 0.0467, "step": 12410 }, { "epoch": 3.85, "learning_rate": 5.0390078015603125e-06, "loss": 0.049, "step": 12411 }, { "epoch": 3.85, "learning_rate": 5.03860772154431e-06, "loss": 0.0553, "step": 12412 }, { "epoch": 3.85, "learning_rate": 5.038207641528307e-06, "loss": 0.023, "step": 12413 }, { "epoch": 3.85, "learning_rate": 5.0378075615123025e-06, "loss": 0.125, "step": 12414 }, { "epoch": 3.85, "learning_rate": 5.037407481496299e-06, "loss": 0.0519, "step": 12415 }, { "epoch": 3.85, "learning_rate": 5.037007401480296e-06, "loss": 0.0319, "step": 12416 }, { "epoch": 3.85, "learning_rate": 5.036607321464293e-06, "loss": 0.009, "step": 12417 }, { "epoch": 3.85, "learning_rate": 5.03620724144829e-06, "loss": 0.0681, "step": 12418 }, { "epoch": 3.85, "learning_rate": 5.035807161432287e-06, "loss": 0.1036, "step": 12419 }, { "epoch": 3.85, "learning_rate": 5.0354070814162835e-06, "loss": 0.0531, "step": 12420 }, { "epoch": 3.86, "learning_rate": 5.03500700140028e-06, "loss": 0.0362, "step": 12421 }, { "epoch": 3.86, "learning_rate": 5.034606921384278e-06, "loss": 0.0164, "step": 12422 }, { "epoch": 3.86, "learning_rate": 5.034206841368274e-06, "loss": 0.0502, "step": 12423 }, { "epoch": 3.86, "learning_rate": 5.033806761352271e-06, "loss": 0.0183, "step": 12424 }, { "epoch": 3.86, "learning_rate": 5.033406681336268e-06, "loss": 0.0451, "step": 12425 }, { "epoch": 3.86, "learning_rate": 5.033006601320265e-06, "loss": 0.0445, "step": 12426 }, { "epoch": 3.86, "learning_rate": 5.032606521304261e-06, "loss": 0.098, "step": 12427 }, { "epoch": 3.86, "learning_rate": 5.032206441288258e-06, "loss": 0.0227, "step": 12428 }, { "epoch": 3.86, "learning_rate": 5.0318063612722545e-06, "loss": 0.0229, "step": 12429 }, { "epoch": 3.86, "learning_rate": 5.031406281256251e-06, "loss": 0.0402, "step": 12430 }, { "epoch": 3.86, "learning_rate": 5.031006201240248e-06, "loss": 0.0434, "step": 12431 }, { "epoch": 3.86, "learning_rate": 5.0306061212242454e-06, "loss": 0.0643, "step": 12432 }, { "epoch": 3.86, "learning_rate": 5.030206041208242e-06, "loss": 0.0447, "step": 12433 }, { "epoch": 3.86, "learning_rate": 5.029805961192239e-06, "loss": 0.0582, "step": 12434 }, { "epoch": 3.86, "learning_rate": 5.0294058811762355e-06, "loss": 0.0363, "step": 12435 }, { "epoch": 3.86, "learning_rate": 5.029005801160233e-06, "loss": 0.0488, "step": 12436 }, { "epoch": 3.86, "learning_rate": 5.02860572114423e-06, "loss": 0.0588, "step": 12437 }, { "epoch": 3.86, "learning_rate": 5.028205641128226e-06, "loss": 0.1128, "step": 12438 }, { "epoch": 3.86, "learning_rate": 5.027805561112223e-06, "loss": 0.064, "step": 12439 }, { "epoch": 3.86, "learning_rate": 5.027405481096219e-06, "loss": 0.0479, "step": 12440 }, { "epoch": 3.86, "learning_rate": 5.0270054010802165e-06, "loss": 0.0223, "step": 12441 }, { "epoch": 3.86, "learning_rate": 5.026605321064213e-06, "loss": 0.0699, "step": 12442 }, { "epoch": 3.86, "learning_rate": 5.02620524104821e-06, "loss": 0.0431, "step": 12443 }, { "epoch": 3.86, "learning_rate": 5.0258051610322065e-06, "loss": 0.0524, "step": 12444 }, { "epoch": 3.86, "learning_rate": 5.025405081016203e-06, "loss": 0.0511, "step": 12445 }, { "epoch": 3.86, "learning_rate": 5.025005001000201e-06, "loss": 0.0281, "step": 12446 }, { "epoch": 3.86, "learning_rate": 5.0246049209841974e-06, "loss": 0.0609, "step": 12447 }, { "epoch": 3.86, "learning_rate": 5.024204840968194e-06, "loss": 0.0199, "step": 12448 }, { "epoch": 3.86, "learning_rate": 5.023804760952191e-06, "loss": 0.021, "step": 12449 }, { "epoch": 3.86, "learning_rate": 5.023404680936188e-06, "loss": 0.0178, "step": 12450 }, { "epoch": 3.86, "learning_rate": 5.023004600920185e-06, "loss": 0.0554, "step": 12451 }, { "epoch": 3.86, "learning_rate": 5.022604520904181e-06, "loss": 0.0256, "step": 12452 }, { "epoch": 3.87, "learning_rate": 5.0222044408881775e-06, "loss": 0.1921, "step": 12453 }, { "epoch": 3.87, "learning_rate": 5.021804360872174e-06, "loss": 0.0266, "step": 12454 }, { "epoch": 3.87, "learning_rate": 5.021404280856172e-06, "loss": 0.0381, "step": 12455 }, { "epoch": 3.87, "learning_rate": 5.0210042008401685e-06, "loss": 0.1163, "step": 12456 }, { "epoch": 3.87, "learning_rate": 5.020604120824165e-06, "loss": 0.0285, "step": 12457 }, { "epoch": 3.87, "learning_rate": 5.020204040808162e-06, "loss": 0.0371, "step": 12458 }, { "epoch": 3.87, "learning_rate": 5.0198039607921585e-06, "loss": 0.0599, "step": 12459 }, { "epoch": 3.87, "learning_rate": 5.019403880776156e-06, "loss": 0.0438, "step": 12460 }, { "epoch": 3.87, "learning_rate": 5.019003800760153e-06, "loss": 0.0293, "step": 12461 }, { "epoch": 3.87, "learning_rate": 5.018603720744149e-06, "loss": 0.0078, "step": 12462 }, { "epoch": 3.87, "learning_rate": 5.018203640728146e-06, "loss": 0.0387, "step": 12463 }, { "epoch": 3.87, "learning_rate": 5.017803560712144e-06, "loss": 0.0629, "step": 12464 }, { "epoch": 3.87, "learning_rate": 5.0174034806961395e-06, "loss": 0.0576, "step": 12465 }, { "epoch": 3.87, "learning_rate": 5.017003400680136e-06, "loss": 0.0471, "step": 12466 }, { "epoch": 3.87, "learning_rate": 5.016603320664133e-06, "loss": 0.0514, "step": 12467 }, { "epoch": 3.87, "learning_rate": 5.0162032406481295e-06, "loss": 0.0193, "step": 12468 }, { "epoch": 3.87, "learning_rate": 5.015803160632127e-06, "loss": 0.0261, "step": 12469 }, { "epoch": 3.87, "learning_rate": 5.015403080616124e-06, "loss": 0.0093, "step": 12470 }, { "epoch": 3.87, "learning_rate": 5.0150030006001204e-06, "loss": 0.0939, "step": 12471 }, { "epoch": 3.87, "learning_rate": 5.014602920584117e-06, "loss": 0.0282, "step": 12472 }, { "epoch": 3.87, "learning_rate": 5.014202840568115e-06, "loss": 0.0413, "step": 12473 }, { "epoch": 3.87, "learning_rate": 5.013802760552111e-06, "loss": 0.0606, "step": 12474 }, { "epoch": 3.87, "learning_rate": 5.013402680536108e-06, "loss": 0.0371, "step": 12475 }, { "epoch": 3.87, "learning_rate": 5.013002600520105e-06, "loss": 0.0312, "step": 12476 }, { "epoch": 3.87, "learning_rate": 5.0126025205041006e-06, "loss": 0.0373, "step": 12477 }, { "epoch": 3.87, "learning_rate": 5.012202440488097e-06, "loss": 0.0177, "step": 12478 }, { "epoch": 3.87, "learning_rate": 5.011802360472095e-06, "loss": 0.0569, "step": 12479 }, { "epoch": 3.87, "learning_rate": 5.0114022804560915e-06, "loss": 0.0581, "step": 12480 }, { "epoch": 3.87, "learning_rate": 5.011002200440088e-06, "loss": 0.0352, "step": 12481 }, { "epoch": 3.87, "learning_rate": 5.010602120424085e-06, "loss": 0.0316, "step": 12482 }, { "epoch": 3.87, "learning_rate": 5.010202040408082e-06, "loss": 0.0558, "step": 12483 }, { "epoch": 3.87, "learning_rate": 5.009801960392079e-06, "loss": 0.0486, "step": 12484 }, { "epoch": 3.88, "learning_rate": 5.009401880376076e-06, "loss": 0.0356, "step": 12485 }, { "epoch": 3.88, "learning_rate": 5.0090018003600724e-06, "loss": 0.0347, "step": 12486 }, { "epoch": 3.88, "learning_rate": 5.00860172034407e-06, "loss": 0.0608, "step": 12487 }, { "epoch": 3.88, "learning_rate": 5.008201640328067e-06, "loss": 0.0136, "step": 12488 }, { "epoch": 3.88, "learning_rate": 5.007801560312063e-06, "loss": 0.0113, "step": 12489 }, { "epoch": 3.88, "learning_rate": 5.007401480296059e-06, "loss": 0.0751, "step": 12490 }, { "epoch": 3.88, "learning_rate": 5.007001400280056e-06, "loss": 0.0373, "step": 12491 }, { "epoch": 3.88, "learning_rate": 5.0066013202640525e-06, "loss": 0.0548, "step": 12492 }, { "epoch": 3.88, "learning_rate": 5.00620124024805e-06, "loss": 0.0611, "step": 12493 }, { "epoch": 3.88, "learning_rate": 5.005801160232047e-06, "loss": 0.039, "step": 12494 }, { "epoch": 3.88, "learning_rate": 5.0054010802160435e-06, "loss": 0.0136, "step": 12495 }, { "epoch": 3.88, "learning_rate": 5.00500100020004e-06, "loss": 0.0907, "step": 12496 }, { "epoch": 3.88, "learning_rate": 5.004600920184038e-06, "loss": 0.0562, "step": 12497 }, { "epoch": 3.88, "learning_rate": 5.004200840168034e-06, "loss": 0.0225, "step": 12498 }, { "epoch": 3.88, "learning_rate": 5.003800760152031e-06, "loss": 0.0389, "step": 12499 }, { "epoch": 3.88, "learning_rate": 5.003400680136028e-06, "loss": 0.0778, "step": 12500 }, { "epoch": 3.88, "eval_loss": 0.1413513422012329, "eval_runtime": 7266.3932, "eval_samples_per_second": 1.504, "eval_steps_per_second": 1.504, "eval_wer": 84.69332245342038, "step": 12500 }, { "epoch": 3.88, "learning_rate": 5.003000600120025e-06, "loss": 0.0327, "step": 12501 }, { "epoch": 3.88, "learning_rate": 5.002600520104022e-06, "loss": 0.0735, "step": 12502 }, { "epoch": 3.88, "learning_rate": 5.002200440088018e-06, "loss": 0.0828, "step": 12503 }, { "epoch": 3.88, "learning_rate": 5.0018003600720145e-06, "loss": 0.0611, "step": 12504 }, { "epoch": 3.88, "learning_rate": 5.001400280056011e-06, "loss": 0.0538, "step": 12505 }, { "epoch": 3.88, "learning_rate": 5.001000200040008e-06, "loss": 0.0467, "step": 12506 }, { "epoch": 3.88, "learning_rate": 5.000600120024005e-06, "loss": 0.1223, "step": 12507 }, { "epoch": 3.88, "learning_rate": 5.000200040008002e-06, "loss": 0.0234, "step": 12508 }, { "epoch": 3.88, "learning_rate": 4.999799959991999e-06, "loss": 0.0572, "step": 12509 }, { "epoch": 3.88, "learning_rate": 4.9993998799759954e-06, "loss": 0.0553, "step": 12510 }, { "epoch": 3.88, "learning_rate": 4.998999799959993e-06, "loss": 0.035, "step": 12511 }, { "epoch": 3.88, "learning_rate": 4.998599719943989e-06, "loss": 0.0394, "step": 12512 }, { "epoch": 3.88, "learning_rate": 4.9981996399279855e-06, "loss": 0.0335, "step": 12513 }, { "epoch": 3.88, "learning_rate": 4.997799559911983e-06, "loss": 0.0369, "step": 12514 }, { "epoch": 3.88, "learning_rate": 4.99739947989598e-06, "loss": 0.0191, "step": 12515 }, { "epoch": 3.88, "learning_rate": 4.996999399879976e-06, "loss": 0.0838, "step": 12516 }, { "epoch": 3.89, "learning_rate": 4.996599319863973e-06, "loss": 0.1515, "step": 12517 }, { "epoch": 3.89, "learning_rate": 4.99619923984797e-06, "loss": 0.0509, "step": 12518 }, { "epoch": 3.89, "learning_rate": 4.9957991598319665e-06, "loss": 0.0525, "step": 12519 }, { "epoch": 3.89, "learning_rate": 4.995399079815963e-06, "loss": 0.0211, "step": 12520 }, { "epoch": 3.89, "learning_rate": 4.994998999799961e-06, "loss": 0.049, "step": 12521 }, { "epoch": 3.89, "learning_rate": 4.994598919783957e-06, "loss": 0.0336, "step": 12522 }, { "epoch": 3.89, "learning_rate": 4.994198839767954e-06, "loss": 0.0745, "step": 12523 }, { "epoch": 3.89, "learning_rate": 4.993798759751951e-06, "loss": 0.0263, "step": 12524 }, { "epoch": 3.89, "learning_rate": 4.9933986797359474e-06, "loss": 0.0336, "step": 12525 }, { "epoch": 3.89, "learning_rate": 4.992998599719944e-06, "loss": 0.1046, "step": 12526 }, { "epoch": 3.89, "learning_rate": 4.992598519703941e-06, "loss": 0.0393, "step": 12527 }, { "epoch": 3.89, "learning_rate": 4.992198439687938e-06, "loss": 0.0226, "step": 12528 }, { "epoch": 3.89, "learning_rate": 4.991798359671935e-06, "loss": 0.0463, "step": 12529 }, { "epoch": 3.89, "learning_rate": 4.991398279655932e-06, "loss": 0.0387, "step": 12530 }, { "epoch": 3.89, "learning_rate": 4.990998199639928e-06, "loss": 0.0182, "step": 12531 }, { "epoch": 3.89, "learning_rate": 4.990598119623925e-06, "loss": 0.0565, "step": 12532 }, { "epoch": 3.89, "learning_rate": 4.990198039607922e-06, "loss": 0.0555, "step": 12533 }, { "epoch": 3.89, "learning_rate": 4.9897979595919185e-06, "loss": 0.0234, "step": 12534 }, { "epoch": 3.89, "learning_rate": 4.989397879575916e-06, "loss": 0.0309, "step": 12535 }, { "epoch": 3.89, "learning_rate": 4.988997799559913e-06, "loss": 0.047, "step": 12536 }, { "epoch": 3.89, "learning_rate": 4.9885977195439085e-06, "loss": 0.054, "step": 12537 }, { "epoch": 3.89, "learning_rate": 4.988197639527906e-06, "loss": 0.0508, "step": 12538 }, { "epoch": 3.89, "learning_rate": 4.987797559511903e-06, "loss": 0.0203, "step": 12539 }, { "epoch": 3.89, "learning_rate": 4.987397479495899e-06, "loss": 0.0277, "step": 12540 }, { "epoch": 3.89, "learning_rate": 4.986997399479896e-06, "loss": 0.0603, "step": 12541 }, { "epoch": 3.89, "learning_rate": 4.986597319463894e-06, "loss": 0.0267, "step": 12542 }, { "epoch": 3.89, "learning_rate": 4.98619723944789e-06, "loss": 0.033, "step": 12543 }, { "epoch": 3.89, "learning_rate": 4.985797159431886e-06, "loss": 0.0578, "step": 12544 }, { "epoch": 3.89, "learning_rate": 4.985397079415884e-06, "loss": 0.03, "step": 12545 }, { "epoch": 3.89, "learning_rate": 4.98499699939988e-06, "loss": 0.0294, "step": 12546 }, { "epoch": 3.89, "learning_rate": 4.984596919383877e-06, "loss": 0.0883, "step": 12547 }, { "epoch": 3.89, "learning_rate": 4.984196839367874e-06, "loss": 0.0245, "step": 12548 }, { "epoch": 3.89, "learning_rate": 4.983796759351871e-06, "loss": 0.0511, "step": 12549 }, { "epoch": 3.9, "learning_rate": 4.983396679335867e-06, "loss": 0.0515, "step": 12550 }, { "epoch": 3.9, "learning_rate": 4.982996599319865e-06, "loss": 0.1428, "step": 12551 }, { "epoch": 3.9, "learning_rate": 4.982596519303861e-06, "loss": 0.0142, "step": 12552 }, { "epoch": 3.9, "learning_rate": 4.982196439287858e-06, "loss": 0.039, "step": 12553 }, { "epoch": 3.9, "learning_rate": 4.981796359271855e-06, "loss": 0.1006, "step": 12554 }, { "epoch": 3.9, "learning_rate": 4.981396279255851e-06, "loss": 0.0361, "step": 12555 }, { "epoch": 3.9, "learning_rate": 4.980996199239848e-06, "loss": 0.0497, "step": 12556 }, { "epoch": 3.9, "learning_rate": 4.980596119223845e-06, "loss": 0.0462, "step": 12557 }, { "epoch": 3.9, "learning_rate": 4.980196039207842e-06, "loss": 0.0343, "step": 12558 }, { "epoch": 3.9, "learning_rate": 4.979795959191839e-06, "loss": 0.0394, "step": 12559 }, { "epoch": 3.9, "learning_rate": 4.979395879175836e-06, "loss": 0.0221, "step": 12560 }, { "epoch": 3.9, "learning_rate": 4.978995799159832e-06, "loss": 0.0291, "step": 12561 }, { "epoch": 3.9, "learning_rate": 4.978595719143829e-06, "loss": 0.0486, "step": 12562 }, { "epoch": 3.9, "learning_rate": 4.978195639127826e-06, "loss": 0.0608, "step": 12563 }, { "epoch": 3.9, "learning_rate": 4.9777955591118224e-06, "loss": 0.0417, "step": 12564 }, { "epoch": 3.9, "learning_rate": 4.97739547909582e-06, "loss": 0.0341, "step": 12565 }, { "epoch": 3.9, "learning_rate": 4.976995399079817e-06, "loss": 0.0542, "step": 12566 }, { "epoch": 3.9, "learning_rate": 4.976595319063813e-06, "loss": 0.0208, "step": 12567 }, { "epoch": 3.9, "learning_rate": 4.97619523904781e-06, "loss": 0.05, "step": 12568 }, { "epoch": 3.9, "learning_rate": 4.975795159031807e-06, "loss": 0.0527, "step": 12569 }, { "epoch": 3.9, "learning_rate": 4.975395079015803e-06, "loss": 0.0456, "step": 12570 }, { "epoch": 3.9, "learning_rate": 4.9749949989998e-06, "loss": 0.0211, "step": 12571 }, { "epoch": 3.9, "learning_rate": 4.974594918983798e-06, "loss": 0.0338, "step": 12572 }, { "epoch": 3.9, "learning_rate": 4.974194838967794e-06, "loss": 0.0339, "step": 12573 }, { "epoch": 3.9, "learning_rate": 4.973794758951791e-06, "loss": 0.0141, "step": 12574 }, { "epoch": 3.9, "learning_rate": 4.973394678935788e-06, "loss": 0.0567, "step": 12575 }, { "epoch": 3.9, "learning_rate": 4.972994598919784e-06, "loss": 0.0494, "step": 12576 }, { "epoch": 3.9, "learning_rate": 4.972594518903781e-06, "loss": 0.1022, "step": 12577 }, { "epoch": 3.9, "learning_rate": 4.972194438887778e-06, "loss": 0.0413, "step": 12578 }, { "epoch": 3.9, "learning_rate": 4.971794358871775e-06, "loss": 0.1044, "step": 12579 }, { "epoch": 3.9, "learning_rate": 4.971394278855772e-06, "loss": 0.0408, "step": 12580 }, { "epoch": 3.9, "learning_rate": 4.970994198839769e-06, "loss": 0.1213, "step": 12581 }, { "epoch": 3.91, "learning_rate": 4.970594118823765e-06, "loss": 0.0336, "step": 12582 }, { "epoch": 3.91, "learning_rate": 4.970194038807762e-06, "loss": 0.0629, "step": 12583 }, { "epoch": 3.91, "learning_rate": 4.969793958791759e-06, "loss": 0.0557, "step": 12584 }, { "epoch": 3.91, "learning_rate": 4.969393878775755e-06, "loss": 0.0473, "step": 12585 }, { "epoch": 3.91, "learning_rate": 4.968993798759753e-06, "loss": 0.0576, "step": 12586 }, { "epoch": 3.91, "learning_rate": 4.96859371874375e-06, "loss": 0.0129, "step": 12587 }, { "epoch": 3.91, "learning_rate": 4.9681936387277454e-06, "loss": 0.0343, "step": 12588 }, { "epoch": 3.91, "learning_rate": 4.967793558711743e-06, "loss": 0.0474, "step": 12589 }, { "epoch": 3.91, "learning_rate": 4.96739347869574e-06, "loss": 0.074, "step": 12590 }, { "epoch": 3.91, "learning_rate": 4.966993398679736e-06, "loss": 0.0265, "step": 12591 }, { "epoch": 3.91, "learning_rate": 4.966593318663733e-06, "loss": 0.024, "step": 12592 }, { "epoch": 3.91, "learning_rate": 4.9661932386477306e-06, "loss": 0.0587, "step": 12593 }, { "epoch": 3.91, "learning_rate": 4.965793158631726e-06, "loss": 0.0504, "step": 12594 }, { "epoch": 3.91, "learning_rate": 4.965393078615723e-06, "loss": 0.0341, "step": 12595 }, { "epoch": 3.91, "learning_rate": 4.964992998599721e-06, "loss": 0.0348, "step": 12596 }, { "epoch": 3.91, "learning_rate": 4.964592918583717e-06, "loss": 0.0978, "step": 12597 }, { "epoch": 3.91, "learning_rate": 4.964192838567714e-06, "loss": 0.0421, "step": 12598 }, { "epoch": 3.91, "learning_rate": 4.963792758551711e-06, "loss": 0.0412, "step": 12599 }, { "epoch": 3.91, "learning_rate": 4.963392678535707e-06, "loss": 0.0467, "step": 12600 }, { "epoch": 3.91, "learning_rate": 4.962992598519704e-06, "loss": 0.0644, "step": 12601 }, { "epoch": 3.91, "learning_rate": 4.962592518503701e-06, "loss": 0.0323, "step": 12602 }, { "epoch": 3.91, "learning_rate": 4.962192438487698e-06, "loss": 0.0399, "step": 12603 }, { "epoch": 3.91, "learning_rate": 4.961792358471695e-06, "loss": 0.0271, "step": 12604 }, { "epoch": 3.91, "learning_rate": 4.961392278455692e-06, "loss": 0.0154, "step": 12605 }, { "epoch": 3.91, "learning_rate": 4.960992198439688e-06, "loss": 0.0572, "step": 12606 }, { "epoch": 3.91, "learning_rate": 4.960592118423685e-06, "loss": 0.0219, "step": 12607 }, { "epoch": 3.91, "learning_rate": 4.960192038407682e-06, "loss": 0.0156, "step": 12608 }, { "epoch": 3.91, "learning_rate": 4.959791958391678e-06, "loss": 0.0313, "step": 12609 }, { "epoch": 3.91, "learning_rate": 4.959391878375676e-06, "loss": 0.0739, "step": 12610 }, { "epoch": 3.91, "learning_rate": 4.958991798359673e-06, "loss": 0.0337, "step": 12611 }, { "epoch": 3.91, "learning_rate": 4.958591718343669e-06, "loss": 0.052, "step": 12612 }, { "epoch": 3.91, "learning_rate": 4.958191638327666e-06, "loss": 0.0379, "step": 12613 }, { "epoch": 3.92, "learning_rate": 4.957791558311663e-06, "loss": 0.0177, "step": 12614 }, { "epoch": 3.92, "learning_rate": 4.957391478295659e-06, "loss": 0.0568, "step": 12615 }, { "epoch": 3.92, "learning_rate": 4.956991398279656e-06, "loss": 0.0304, "step": 12616 }, { "epoch": 3.92, "learning_rate": 4.9565913182636536e-06, "loss": 0.0268, "step": 12617 }, { "epoch": 3.92, "learning_rate": 4.95619123824765e-06, "loss": 0.0386, "step": 12618 }, { "epoch": 3.92, "learning_rate": 4.955791158231646e-06, "loss": 0.0329, "step": 12619 }, { "epoch": 3.92, "learning_rate": 4.955391078215644e-06, "loss": 0.0324, "step": 12620 }, { "epoch": 3.92, "learning_rate": 4.95499099819964e-06, "loss": 0.0339, "step": 12621 }, { "epoch": 3.92, "learning_rate": 4.954590918183637e-06, "loss": 0.0251, "step": 12622 }, { "epoch": 3.92, "learning_rate": 4.954190838167634e-06, "loss": 0.0692, "step": 12623 }, { "epoch": 3.92, "learning_rate": 4.953790758151631e-06, "loss": 0.052, "step": 12624 }, { "epoch": 3.92, "learning_rate": 4.953390678135628e-06, "loss": 0.1029, "step": 12625 }, { "epoch": 3.92, "learning_rate": 4.952990598119624e-06, "loss": 0.0363, "step": 12626 }, { "epoch": 3.92, "learning_rate": 4.952590518103621e-06, "loss": 0.0389, "step": 12627 }, { "epoch": 3.92, "learning_rate": 4.952190438087618e-06, "loss": 0.0734, "step": 12628 }, { "epoch": 3.92, "learning_rate": 4.951790358071615e-06, "loss": 0.0149, "step": 12629 }, { "epoch": 3.92, "learning_rate": 4.951390278055611e-06, "loss": 0.0449, "step": 12630 }, { "epoch": 3.92, "learning_rate": 4.950990198039609e-06, "loss": 0.0272, "step": 12631 }, { "epoch": 3.92, "learning_rate": 4.950590118023605e-06, "loss": 0.0175, "step": 12632 }, { "epoch": 3.92, "learning_rate": 4.950190038007601e-06, "loss": 0.0211, "step": 12633 }, { "epoch": 3.92, "learning_rate": 4.949789957991599e-06, "loss": 0.0382, "step": 12634 }, { "epoch": 3.92, "learning_rate": 4.949389877975596e-06, "loss": 0.0836, "step": 12635 }, { "epoch": 3.92, "learning_rate": 4.948989797959592e-06, "loss": 0.0713, "step": 12636 }, { "epoch": 3.92, "learning_rate": 4.948589717943589e-06, "loss": 0.0701, "step": 12637 }, { "epoch": 3.92, "learning_rate": 4.948189637927586e-06, "loss": 0.0273, "step": 12638 }, { "epoch": 3.92, "learning_rate": 4.947789557911582e-06, "loss": 0.0527, "step": 12639 }, { "epoch": 3.92, "learning_rate": 4.947389477895579e-06, "loss": 0.0333, "step": 12640 }, { "epoch": 3.92, "learning_rate": 4.946989397879577e-06, "loss": 0.0643, "step": 12641 }, { "epoch": 3.92, "learning_rate": 4.946589317863573e-06, "loss": 0.0524, "step": 12642 }, { "epoch": 3.92, "learning_rate": 4.94618923784757e-06, "loss": 0.0499, "step": 12643 }, { "epoch": 3.92, "learning_rate": 4.945789157831567e-06, "loss": 0.0306, "step": 12644 }, { "epoch": 3.92, "learning_rate": 4.945389077815563e-06, "loss": 0.1423, "step": 12645 }, { "epoch": 3.93, "learning_rate": 4.94498899779956e-06, "loss": 0.0491, "step": 12646 }, { "epoch": 3.93, "learning_rate": 4.944588917783557e-06, "loss": 0.0986, "step": 12647 }, { "epoch": 3.93, "learning_rate": 4.944188837767554e-06, "loss": 0.0365, "step": 12648 }, { "epoch": 3.93, "learning_rate": 4.943788757751551e-06, "loss": 0.0132, "step": 12649 }, { "epoch": 3.93, "learning_rate": 4.943388677735548e-06, "loss": 0.0575, "step": 12650 }, { "epoch": 3.93, "learning_rate": 4.942988597719544e-06, "loss": 0.0408, "step": 12651 }, { "epoch": 3.93, "learning_rate": 4.942588517703541e-06, "loss": 0.0557, "step": 12652 }, { "epoch": 3.93, "learning_rate": 4.942188437687538e-06, "loss": 0.0185, "step": 12653 }, { "epoch": 3.93, "learning_rate": 4.941788357671534e-06, "loss": 0.0228, "step": 12654 }, { "epoch": 3.93, "learning_rate": 4.941388277655532e-06, "loss": 0.03, "step": 12655 }, { "epoch": 3.93, "learning_rate": 4.9409881976395286e-06, "loss": 0.1659, "step": 12656 }, { "epoch": 3.93, "learning_rate": 4.940588117623525e-06, "loss": 0.0771, "step": 12657 }, { "epoch": 3.93, "learning_rate": 4.940188037607522e-06, "loss": 0.0349, "step": 12658 }, { "epoch": 3.93, "learning_rate": 4.939787957591519e-06, "loss": 0.0727, "step": 12659 }, { "epoch": 3.93, "learning_rate": 4.939387877575515e-06, "loss": 0.0371, "step": 12660 }, { "epoch": 3.93, "learning_rate": 4.938987797559512e-06, "loss": 0.0418, "step": 12661 }, { "epoch": 3.93, "learning_rate": 4.9385877175435095e-06, "loss": 0.0836, "step": 12662 }, { "epoch": 3.93, "learning_rate": 4.938187637527505e-06, "loss": 0.0677, "step": 12663 }, { "epoch": 3.93, "learning_rate": 4.937787557511503e-06, "loss": 0.1148, "step": 12664 }, { "epoch": 3.93, "learning_rate": 4.9373874774955e-06, "loss": 0.0348, "step": 12665 }, { "epoch": 3.93, "learning_rate": 4.936987397479496e-06, "loss": 0.0336, "step": 12666 }, { "epoch": 3.93, "learning_rate": 4.936587317463493e-06, "loss": 0.0671, "step": 12667 }, { "epoch": 3.93, "learning_rate": 4.93618723744749e-06, "loss": 0.0348, "step": 12668 }, { "epoch": 3.93, "learning_rate": 4.935787157431487e-06, "loss": 0.0489, "step": 12669 }, { "epoch": 3.93, "learning_rate": 4.935387077415483e-06, "loss": 0.0327, "step": 12670 }, { "epoch": 3.93, "learning_rate": 4.9349869973994806e-06, "loss": 0.0374, "step": 12671 }, { "epoch": 3.93, "learning_rate": 4.934586917383477e-06, "loss": 0.0683, "step": 12672 }, { "epoch": 3.93, "learning_rate": 4.934186837367474e-06, "loss": 0.0763, "step": 12673 }, { "epoch": 3.93, "learning_rate": 4.933786757351471e-06, "loss": 0.1399, "step": 12674 }, { "epoch": 3.93, "learning_rate": 4.933386677335467e-06, "loss": 0.0584, "step": 12675 }, { "epoch": 3.93, "learning_rate": 4.932986597319464e-06, "loss": 0.0586, "step": 12676 }, { "epoch": 3.93, "learning_rate": 4.932586517303461e-06, "loss": 0.0434, "step": 12677 }, { "epoch": 3.93, "learning_rate": 4.932186437287458e-06, "loss": 0.0299, "step": 12678 }, { "epoch": 3.94, "learning_rate": 4.931786357271455e-06, "loss": 0.043, "step": 12679 }, { "epoch": 3.94, "learning_rate": 4.931386277255452e-06, "loss": 0.015, "step": 12680 }, { "epoch": 3.94, "learning_rate": 4.930986197239448e-06, "loss": 0.0252, "step": 12681 }, { "epoch": 3.94, "learning_rate": 4.930586117223445e-06, "loss": 0.0272, "step": 12682 }, { "epoch": 3.94, "learning_rate": 4.930186037207442e-06, "loss": 0.0253, "step": 12683 }, { "epoch": 3.94, "learning_rate": 4.929785957191438e-06, "loss": 0.0095, "step": 12684 }, { "epoch": 3.94, "learning_rate": 4.929385877175436e-06, "loss": 0.0328, "step": 12685 }, { "epoch": 3.94, "learning_rate": 4.9289857971594326e-06, "loss": 0.0313, "step": 12686 }, { "epoch": 3.94, "learning_rate": 4.928585717143429e-06, "loss": 0.0167, "step": 12687 }, { "epoch": 3.94, "learning_rate": 4.928185637127426e-06, "loss": 0.0286, "step": 12688 }, { "epoch": 3.94, "learning_rate": 4.927785557111423e-06, "loss": 0.0663, "step": 12689 }, { "epoch": 3.94, "learning_rate": 4.927385477095419e-06, "loss": 0.0548, "step": 12690 }, { "epoch": 3.94, "learning_rate": 4.926985397079416e-06, "loss": 0.0469, "step": 12691 }, { "epoch": 3.94, "learning_rate": 4.9265853170634135e-06, "loss": 0.0699, "step": 12692 }, { "epoch": 3.94, "learning_rate": 4.92618523704741e-06, "loss": 0.0312, "step": 12693 }, { "epoch": 3.94, "learning_rate": 4.925785157031407e-06, "loss": 0.0573, "step": 12694 }, { "epoch": 3.94, "learning_rate": 4.9253850770154036e-06, "loss": 0.0306, "step": 12695 }, { "epoch": 3.94, "learning_rate": 4.9249849969994e-06, "loss": 0.0496, "step": 12696 }, { "epoch": 3.94, "learning_rate": 4.924584916983397e-06, "loss": 0.1331, "step": 12697 }, { "epoch": 3.94, "learning_rate": 4.924184836967394e-06, "loss": 0.0275, "step": 12698 }, { "epoch": 3.94, "learning_rate": 4.923784756951391e-06, "loss": 0.0176, "step": 12699 }, { "epoch": 3.94, "learning_rate": 4.923384676935388e-06, "loss": 0.0212, "step": 12700 }, { "epoch": 3.94, "learning_rate": 4.922984596919384e-06, "loss": 0.0425, "step": 12701 }, { "epoch": 3.94, "learning_rate": 4.922584516903381e-06, "loss": 0.026, "step": 12702 }, { "epoch": 3.94, "learning_rate": 4.922184436887378e-06, "loss": 0.0411, "step": 12703 }, { "epoch": 3.94, "learning_rate": 4.921784356871375e-06, "loss": 0.0397, "step": 12704 }, { "epoch": 3.94, "learning_rate": 4.921384276855371e-06, "loss": 0.027, "step": 12705 }, { "epoch": 3.94, "learning_rate": 4.920984196839369e-06, "loss": 0.0224, "step": 12706 }, { "epoch": 3.94, "learning_rate": 4.9205841168233655e-06, "loss": 0.0326, "step": 12707 }, { "epoch": 3.94, "learning_rate": 4.920184036807361e-06, "loss": 0.0175, "step": 12708 }, { "epoch": 3.94, "learning_rate": 4.919783956791359e-06, "loss": 0.0347, "step": 12709 }, { "epoch": 3.94, "learning_rate": 4.9193838767753556e-06, "loss": 0.028, "step": 12710 }, { "epoch": 3.95, "learning_rate": 4.918983796759352e-06, "loss": 0.0408, "step": 12711 }, { "epoch": 3.95, "learning_rate": 4.918583716743349e-06, "loss": 0.0691, "step": 12712 }, { "epoch": 3.95, "learning_rate": 4.9181836367273465e-06, "loss": 0.0231, "step": 12713 }, { "epoch": 3.95, "learning_rate": 4.917783556711342e-06, "loss": 0.0519, "step": 12714 }, { "epoch": 3.95, "learning_rate": 4.917383476695339e-06, "loss": 0.0447, "step": 12715 }, { "epoch": 3.95, "learning_rate": 4.9169833966793365e-06, "loss": 0.1159, "step": 12716 }, { "epoch": 3.95, "learning_rate": 4.916583316663333e-06, "loss": 0.0433, "step": 12717 }, { "epoch": 3.95, "learning_rate": 4.91618323664733e-06, "loss": 0.0418, "step": 12718 }, { "epoch": 3.95, "learning_rate": 4.915783156631327e-06, "loss": 0.0415, "step": 12719 }, { "epoch": 3.95, "learning_rate": 4.915383076615323e-06, "loss": 0.1287, "step": 12720 }, { "epoch": 3.95, "learning_rate": 4.91498299659932e-06, "loss": 0.0155, "step": 12721 }, { "epoch": 3.95, "learning_rate": 4.914582916583317e-06, "loss": 0.0332, "step": 12722 }, { "epoch": 3.95, "learning_rate": 4.914182836567314e-06, "loss": 0.055, "step": 12723 }, { "epoch": 3.95, "learning_rate": 4.913782756551311e-06, "loss": 0.0473, "step": 12724 }, { "epoch": 3.95, "learning_rate": 4.9133826765353076e-06, "loss": 0.023, "step": 12725 }, { "epoch": 3.95, "learning_rate": 4.912982596519304e-06, "loss": 0.0276, "step": 12726 }, { "epoch": 3.95, "learning_rate": 4.912582516503301e-06, "loss": 0.024, "step": 12727 }, { "epoch": 3.95, "learning_rate": 4.912182436487298e-06, "loss": 0.0144, "step": 12728 }, { "epoch": 3.95, "learning_rate": 4.911782356471294e-06, "loss": 0.0402, "step": 12729 }, { "epoch": 3.95, "learning_rate": 4.911382276455292e-06, "loss": 0.0644, "step": 12730 }, { "epoch": 3.95, "learning_rate": 4.9109821964392885e-06, "loss": 0.0084, "step": 12731 }, { "epoch": 3.95, "learning_rate": 4.910582116423285e-06, "loss": 0.0387, "step": 12732 }, { "epoch": 3.95, "learning_rate": 4.910182036407282e-06, "loss": 0.049, "step": 12733 }, { "epoch": 3.95, "learning_rate": 4.9097819563912786e-06, "loss": 0.0557, "step": 12734 }, { "epoch": 3.95, "learning_rate": 4.909381876375275e-06, "loss": 0.0259, "step": 12735 }, { "epoch": 3.95, "learning_rate": 4.908981796359272e-06, "loss": 0.0492, "step": 12736 }, { "epoch": 3.95, "learning_rate": 4.9085817163432695e-06, "loss": 0.0452, "step": 12737 }, { "epoch": 3.95, "learning_rate": 4.908181636327266e-06, "loss": 0.1073, "step": 12738 }, { "epoch": 3.95, "learning_rate": 4.907781556311262e-06, "loss": 0.0238, "step": 12739 }, { "epoch": 3.95, "learning_rate": 4.9073814762952595e-06, "loss": 0.0346, "step": 12740 }, { "epoch": 3.95, "learning_rate": 4.906981396279256e-06, "loss": 0.0593, "step": 12741 }, { "epoch": 3.95, "learning_rate": 4.906581316263253e-06, "loss": 0.025, "step": 12742 }, { "epoch": 3.96, "learning_rate": 4.90618123624725e-06, "loss": 0.0251, "step": 12743 }, { "epoch": 3.96, "learning_rate": 4.905781156231247e-06, "loss": 0.0464, "step": 12744 }, { "epoch": 3.96, "learning_rate": 4.905381076215243e-06, "loss": 0.0413, "step": 12745 }, { "epoch": 3.96, "learning_rate": 4.90498099619924e-06, "loss": 0.0334, "step": 12746 }, { "epoch": 3.96, "learning_rate": 4.904580916183237e-06, "loss": 0.0505, "step": 12747 }, { "epoch": 3.96, "learning_rate": 4.904180836167234e-06, "loss": 0.0533, "step": 12748 }, { "epoch": 3.96, "learning_rate": 4.9037807561512306e-06, "loss": 0.1219, "step": 12749 }, { "epoch": 3.96, "learning_rate": 4.903380676135227e-06, "loss": 0.014, "step": 12750 }, { "epoch": 3.96, "learning_rate": 4.902980596119225e-06, "loss": 0.0848, "step": 12751 }, { "epoch": 3.96, "learning_rate": 4.902580516103221e-06, "loss": 0.0139, "step": 12752 }, { "epoch": 3.96, "learning_rate": 4.902180436087217e-06, "loss": 0.0644, "step": 12753 }, { "epoch": 3.96, "learning_rate": 4.901780356071215e-06, "loss": 0.0221, "step": 12754 }, { "epoch": 3.96, "learning_rate": 4.9013802760552115e-06, "loss": 0.0869, "step": 12755 }, { "epoch": 3.96, "learning_rate": 4.900980196039208e-06, "loss": 0.0586, "step": 12756 }, { "epoch": 3.96, "learning_rate": 4.900580116023205e-06, "loss": 0.0384, "step": 12757 }, { "epoch": 3.96, "learning_rate": 4.900180036007202e-06, "loss": 0.0379, "step": 12758 }, { "epoch": 3.96, "learning_rate": 4.899779955991198e-06, "loss": 0.0559, "step": 12759 }, { "epoch": 3.96, "learning_rate": 4.899379875975195e-06, "loss": 0.0426, "step": 12760 }, { "epoch": 3.96, "learning_rate": 4.8989797959591925e-06, "loss": 0.0226, "step": 12761 }, { "epoch": 3.96, "learning_rate": 4.898579715943189e-06, "loss": 0.0193, "step": 12762 }, { "epoch": 3.96, "learning_rate": 4.898179635927186e-06, "loss": 0.064, "step": 12763 }, { "epoch": 3.96, "learning_rate": 4.8977795559111826e-06, "loss": 0.0699, "step": 12764 }, { "epoch": 3.96, "learning_rate": 4.897379475895179e-06, "loss": 0.0496, "step": 12765 }, { "epoch": 3.96, "learning_rate": 4.896979395879176e-06, "loss": 0.0369, "step": 12766 }, { "epoch": 3.96, "learning_rate": 4.896579315863173e-06, "loss": 0.0296, "step": 12767 }, { "epoch": 3.96, "learning_rate": 4.89617923584717e-06, "loss": 0.1633, "step": 12768 }, { "epoch": 3.96, "learning_rate": 4.895779155831167e-06, "loss": 0.0136, "step": 12769 }, { "epoch": 3.96, "learning_rate": 4.8953790758151635e-06, "loss": 0.0137, "step": 12770 }, { "epoch": 3.96, "learning_rate": 4.89497899579916e-06, "loss": 0.0451, "step": 12771 }, { "epoch": 3.96, "learning_rate": 4.894578915783157e-06, "loss": 0.0371, "step": 12772 }, { "epoch": 3.96, "learning_rate": 4.8941788357671536e-06, "loss": 0.0526, "step": 12773 }, { "epoch": 3.96, "learning_rate": 4.893778755751151e-06, "loss": 0.0131, "step": 12774 }, { "epoch": 3.97, "learning_rate": 4.893378675735148e-06, "loss": 0.1054, "step": 12775 }, { "epoch": 3.97, "learning_rate": 4.8929785957191445e-06, "loss": 0.0354, "step": 12776 }, { "epoch": 3.97, "learning_rate": 4.892578515703141e-06, "loss": 0.0346, "step": 12777 }, { "epoch": 3.97, "learning_rate": 4.892178435687138e-06, "loss": 0.037, "step": 12778 }, { "epoch": 3.97, "learning_rate": 4.8917783556711345e-06, "loss": 0.1363, "step": 12779 }, { "epoch": 3.97, "learning_rate": 4.891378275655131e-06, "loss": 0.0489, "step": 12780 }, { "epoch": 3.97, "learning_rate": 4.890978195639129e-06, "loss": 0.0748, "step": 12781 }, { "epoch": 3.97, "learning_rate": 4.8905781156231254e-06, "loss": 0.0189, "step": 12782 }, { "epoch": 3.97, "learning_rate": 4.890178035607121e-06, "loss": 0.0436, "step": 12783 }, { "epoch": 3.97, "learning_rate": 4.889777955591119e-06, "loss": 0.0561, "step": 12784 }, { "epoch": 3.97, "learning_rate": 4.8893778755751155e-06, "loss": 0.0239, "step": 12785 }, { "epoch": 3.97, "learning_rate": 4.888977795559112e-06, "loss": 0.0274, "step": 12786 }, { "epoch": 3.97, "learning_rate": 4.888577715543109e-06, "loss": 0.0741, "step": 12787 }, { "epoch": 3.97, "learning_rate": 4.888177635527106e-06, "loss": 0.0608, "step": 12788 }, { "epoch": 3.97, "learning_rate": 4.887777555511103e-06, "loss": 0.015, "step": 12789 }, { "epoch": 3.97, "learning_rate": 4.887377475495099e-06, "loss": 0.1921, "step": 12790 }, { "epoch": 3.97, "learning_rate": 4.8869773954790965e-06, "loss": 0.0299, "step": 12791 }, { "epoch": 3.97, "learning_rate": 4.886577315463093e-06, "loss": 0.0859, "step": 12792 }, { "epoch": 3.97, "learning_rate": 4.88617723544709e-06, "loss": 0.1347, "step": 12793 }, { "epoch": 3.97, "learning_rate": 4.8857771554310865e-06, "loss": 0.0455, "step": 12794 }, { "epoch": 3.97, "learning_rate": 4.885377075415084e-06, "loss": 0.0545, "step": 12795 }, { "epoch": 3.97, "learning_rate": 4.88497699539908e-06, "loss": 0.0584, "step": 12796 }, { "epoch": 3.97, "learning_rate": 4.884576915383077e-06, "loss": 0.0639, "step": 12797 }, { "epoch": 3.97, "learning_rate": 4.884176835367074e-06, "loss": 0.0542, "step": 12798 }, { "epoch": 3.97, "learning_rate": 4.883776755351071e-06, "loss": 0.0208, "step": 12799 }, { "epoch": 3.97, "learning_rate": 4.8833766753350675e-06, "loss": 0.1428, "step": 12800 }, { "epoch": 3.97, "learning_rate": 4.882976595319064e-06, "loss": 0.0293, "step": 12801 }, { "epoch": 3.97, "learning_rate": 4.882576515303061e-06, "loss": 0.0837, "step": 12802 }, { "epoch": 3.97, "learning_rate": 4.8821764352870576e-06, "loss": 0.0523, "step": 12803 }, { "epoch": 3.97, "learning_rate": 4.881776355271054e-06, "loss": 0.0316, "step": 12804 }, { "epoch": 3.97, "learning_rate": 4.881376275255052e-06, "loss": 0.0217, "step": 12805 }, { "epoch": 3.97, "learning_rate": 4.8809761952390485e-06, "loss": 0.0332, "step": 12806 }, { "epoch": 3.98, "learning_rate": 4.880576115223045e-06, "loss": 0.0416, "step": 12807 }, { "epoch": 3.98, "learning_rate": 4.880176035207042e-06, "loss": 0.0221, "step": 12808 }, { "epoch": 3.98, "learning_rate": 4.8797759551910385e-06, "loss": 0.1103, "step": 12809 }, { "epoch": 3.98, "learning_rate": 4.879375875175035e-06, "loss": 0.0345, "step": 12810 }, { "epoch": 3.98, "learning_rate": 4.878975795159032e-06, "loss": 0.0522, "step": 12811 }, { "epoch": 3.98, "learning_rate": 4.878575715143029e-06, "loss": 0.0419, "step": 12812 }, { "epoch": 3.98, "learning_rate": 4.878175635127026e-06, "loss": 0.0338, "step": 12813 }, { "epoch": 3.98, "learning_rate": 4.877775555111023e-06, "loss": 0.0221, "step": 12814 }, { "epoch": 3.98, "learning_rate": 4.8773754750950195e-06, "loss": 0.0652, "step": 12815 }, { "epoch": 3.98, "learning_rate": 4.876975395079016e-06, "loss": 0.0222, "step": 12816 }, { "epoch": 3.98, "learning_rate": 4.876575315063013e-06, "loss": 0.0481, "step": 12817 }, { "epoch": 3.98, "learning_rate": 4.8761752350470095e-06, "loss": 0.0389, "step": 12818 }, { "epoch": 3.98, "learning_rate": 4.875775155031007e-06, "loss": 0.0297, "step": 12819 }, { "epoch": 3.98, "learning_rate": 4.875375075015004e-06, "loss": 0.1158, "step": 12820 }, { "epoch": 3.98, "learning_rate": 4.874974994999e-06, "loss": 0.0389, "step": 12821 }, { "epoch": 3.98, "learning_rate": 4.874574914982997e-06, "loss": 0.0806, "step": 12822 }, { "epoch": 3.98, "learning_rate": 4.874174834966994e-06, "loss": 0.0761, "step": 12823 }, { "epoch": 3.98, "learning_rate": 4.8737747549509905e-06, "loss": 0.0291, "step": 12824 }, { "epoch": 3.98, "learning_rate": 4.873374674934987e-06, "loss": 0.0609, "step": 12825 }, { "epoch": 3.98, "learning_rate": 4.872974594918985e-06, "loss": 0.1229, "step": 12826 }, { "epoch": 3.98, "learning_rate": 4.8725745149029806e-06, "loss": 0.0398, "step": 12827 }, { "epoch": 3.98, "learning_rate": 4.872174434886977e-06, "loss": 0.0447, "step": 12828 }, { "epoch": 3.98, "learning_rate": 4.871774354870975e-06, "loss": 0.0284, "step": 12829 }, { "epoch": 3.98, "learning_rate": 4.8713742748549715e-06, "loss": 0.1243, "step": 12830 }, { "epoch": 3.98, "learning_rate": 4.870974194838968e-06, "loss": 0.0115, "step": 12831 }, { "epoch": 3.98, "learning_rate": 4.870574114822965e-06, "loss": 0.1522, "step": 12832 }, { "epoch": 3.98, "learning_rate": 4.870174034806962e-06, "loss": 0.0937, "step": 12833 }, { "epoch": 3.98, "learning_rate": 4.869773954790958e-06, "loss": 0.0722, "step": 12834 }, { "epoch": 3.98, "learning_rate": 4.869373874774955e-06, "loss": 0.0554, "step": 12835 }, { "epoch": 3.98, "learning_rate": 4.8689737947589524e-06, "loss": 0.042, "step": 12836 }, { "epoch": 3.98, "learning_rate": 4.868573714742949e-06, "loss": 0.0443, "step": 12837 }, { "epoch": 3.98, "learning_rate": 4.868173634726946e-06, "loss": 0.0655, "step": 12838 }, { "epoch": 3.98, "learning_rate": 4.8677735547109425e-06, "loss": 0.1302, "step": 12839 }, { "epoch": 3.99, "learning_rate": 4.867373474694939e-06, "loss": 0.039, "step": 12840 }, { "epoch": 3.99, "learning_rate": 4.866973394678936e-06, "loss": 0.0819, "step": 12841 }, { "epoch": 3.99, "learning_rate": 4.8665733146629326e-06, "loss": 0.0233, "step": 12842 }, { "epoch": 3.99, "learning_rate": 4.86617323464693e-06, "loss": 0.0406, "step": 12843 }, { "epoch": 3.99, "learning_rate": 4.865773154630927e-06, "loss": 0.0415, "step": 12844 }, { "epoch": 3.99, "learning_rate": 4.8653730746149235e-06, "loss": 0.0482, "step": 12845 }, { "epoch": 3.99, "learning_rate": 4.86497299459892e-06, "loss": 0.0627, "step": 12846 }, { "epoch": 3.99, "learning_rate": 4.864572914582917e-06, "loss": 0.0482, "step": 12847 }, { "epoch": 3.99, "learning_rate": 4.8641728345669135e-06, "loss": 0.014, "step": 12848 }, { "epoch": 3.99, "learning_rate": 4.86377275455091e-06, "loss": 0.033, "step": 12849 }, { "epoch": 3.99, "learning_rate": 4.863372674534908e-06, "loss": 0.0133, "step": 12850 }, { "epoch": 3.99, "learning_rate": 4.862972594518904e-06, "loss": 0.0327, "step": 12851 }, { "epoch": 3.99, "learning_rate": 4.862572514502901e-06, "loss": 0.0568, "step": 12852 }, { "epoch": 3.99, "learning_rate": 4.862172434486898e-06, "loss": 0.0388, "step": 12853 }, { "epoch": 3.99, "learning_rate": 4.8617723544708945e-06, "loss": 0.0145, "step": 12854 }, { "epoch": 3.99, "learning_rate": 4.861372274454891e-06, "loss": 0.0363, "step": 12855 }, { "epoch": 3.99, "learning_rate": 4.860972194438888e-06, "loss": 0.032, "step": 12856 }, { "epoch": 3.99, "learning_rate": 4.860572114422885e-06, "loss": 0.0465, "step": 12857 }, { "epoch": 3.99, "learning_rate": 4.860172034406882e-06, "loss": 0.0701, "step": 12858 }, { "epoch": 3.99, "learning_rate": 4.859771954390879e-06, "loss": 0.0151, "step": 12859 }, { "epoch": 3.99, "learning_rate": 4.8593718743748754e-06, "loss": 0.0715, "step": 12860 }, { "epoch": 3.99, "learning_rate": 4.858971794358872e-06, "loss": 0.0214, "step": 12861 }, { "epoch": 3.99, "learning_rate": 4.858571714342869e-06, "loss": 0.1119, "step": 12862 }, { "epoch": 3.99, "learning_rate": 4.8581716343268655e-06, "loss": 0.0873, "step": 12863 }, { "epoch": 3.99, "learning_rate": 4.857771554310863e-06, "loss": 0.0608, "step": 12864 }, { "epoch": 3.99, "learning_rate": 4.857371474294859e-06, "loss": 0.0776, "step": 12865 }, { "epoch": 3.99, "learning_rate": 4.856971394278856e-06, "loss": 0.0335, "step": 12866 }, { "epoch": 3.99, "learning_rate": 4.856571314262853e-06, "loss": 0.0253, "step": 12867 }, { "epoch": 3.99, "learning_rate": 4.85617123424685e-06, "loss": 0.0455, "step": 12868 }, { "epoch": 3.99, "learning_rate": 4.8557711542308465e-06, "loss": 0.0891, "step": 12869 }, { "epoch": 3.99, "learning_rate": 4.855371074214843e-06, "loss": 0.0504, "step": 12870 }, { "epoch": 3.99, "learning_rate": 4.854970994198841e-06, "loss": 0.0198, "step": 12871 }, { "epoch": 4.0, "learning_rate": 4.8545709141828365e-06, "loss": 0.039, "step": 12872 }, { "epoch": 4.0, "learning_rate": 4.854170834166834e-06, "loss": 0.0107, "step": 12873 }, { "epoch": 4.0, "learning_rate": 4.853770754150831e-06, "loss": 0.0224, "step": 12874 }, { "epoch": 4.0, "learning_rate": 4.8533706741348274e-06, "loss": 0.0261, "step": 12875 }, { "epoch": 4.0, "learning_rate": 4.852970594118824e-06, "loss": 0.0328, "step": 12876 }, { "epoch": 4.0, "learning_rate": 4.852570514102821e-06, "loss": 0.0303, "step": 12877 }, { "epoch": 4.0, "learning_rate": 4.8521704340868175e-06, "loss": 0.0754, "step": 12878 }, { "epoch": 4.0, "learning_rate": 4.851770354070814e-06, "loss": 0.1287, "step": 12879 }, { "epoch": 4.0, "learning_rate": 4.851370274054812e-06, "loss": 0.0473, "step": 12880 }, { "epoch": 4.0, "learning_rate": 4.850970194038808e-06, "loss": 0.0237, "step": 12881 }, { "epoch": 4.0, "learning_rate": 4.850570114022805e-06, "loss": 0.0308, "step": 12882 }, { "epoch": 4.0, "learning_rate": 4.850170034006802e-06, "loss": 0.067, "step": 12883 }, { "epoch": 4.0, "learning_rate": 4.8497699539907985e-06, "loss": 0.0399, "step": 12884 }, { "epoch": 4.0, "learning_rate": 4.849369873974795e-06, "loss": 0.0244, "step": 12885 }, { "epoch": 4.0, "learning_rate": 4.848969793958792e-06, "loss": 0.0619, "step": 12886 }, { "epoch": 4.0, "learning_rate": 4.848569713942789e-06, "loss": 0.0467, "step": 12887 }, { "epoch": 4.0, "learning_rate": 4.848169633926786e-06, "loss": 0.0278, "step": 12888 }, { "epoch": 4.0, "learning_rate": 4.847769553910783e-06, "loss": 0.0393, "step": 12889 }, { "epoch": 4.0, "learning_rate": 4.847369473894779e-06, "loss": 0.0841, "step": 12890 }, { "epoch": 4.0, "learning_rate": 4.846969393878776e-06, "loss": 0.0175, "step": 12891 }, { "epoch": 4.0, "learning_rate": 4.846569313862773e-06, "loss": 0.0926, "step": 12892 }, { "epoch": 4.0, "learning_rate": 4.8461692338467695e-06, "loss": 0.009, "step": 12893 }, { "epoch": 4.0, "learning_rate": 4.845769153830767e-06, "loss": 0.0608, "step": 12894 }, { "epoch": 4.0, "learning_rate": 4.845369073814764e-06, "loss": 0.013, "step": 12895 }, { "epoch": 4.0, "learning_rate": 4.84496899379876e-06, "loss": 0.0948, "step": 12896 }, { "epoch": 4.0, "learning_rate": 4.844568913782757e-06, "loss": 0.0575, "step": 12897 }, { "epoch": 4.0, "learning_rate": 4.844168833766754e-06, "loss": 0.0151, "step": 12898 }, { "epoch": 4.0, "learning_rate": 4.8437687537507504e-06, "loss": 0.0182, "step": 12899 }, { "epoch": 4.0, "learning_rate": 4.843368673734747e-06, "loss": 0.0608, "step": 12900 }, { "epoch": 4.0, "learning_rate": 4.842968593718745e-06, "loss": 0.0227, "step": 12901 }, { "epoch": 4.0, "learning_rate": 4.842568513702741e-06, "loss": 0.0325, "step": 12902 }, { "epoch": 4.0, "learning_rate": 4.842168433686737e-06, "loss": 0.0283, "step": 12903 }, { "epoch": 4.01, "learning_rate": 4.841768353670735e-06, "loss": 0.0056, "step": 12904 }, { "epoch": 4.01, "learning_rate": 4.841368273654731e-06, "loss": 0.0189, "step": 12905 }, { "epoch": 4.01, "learning_rate": 4.840968193638728e-06, "loss": 0.0209, "step": 12906 }, { "epoch": 4.01, "learning_rate": 4.840568113622725e-06, "loss": 0.0166, "step": 12907 }, { "epoch": 4.01, "learning_rate": 4.840168033606722e-06, "loss": 0.0138, "step": 12908 }, { "epoch": 4.01, "learning_rate": 4.839767953590718e-06, "loss": 0.0091, "step": 12909 }, { "epoch": 4.01, "learning_rate": 4.839367873574715e-06, "loss": 0.0342, "step": 12910 }, { "epoch": 4.01, "learning_rate": 4.838967793558712e-06, "loss": 0.0063, "step": 12911 }, { "epoch": 4.01, "learning_rate": 4.838567713542709e-06, "loss": 0.0156, "step": 12912 }, { "epoch": 4.01, "learning_rate": 4.838167633526706e-06, "loss": 0.0156, "step": 12913 }, { "epoch": 4.01, "learning_rate": 4.8377675535107024e-06, "loss": 0.0367, "step": 12914 }, { "epoch": 4.01, "learning_rate": 4.8373674734947e-06, "loss": 0.0151, "step": 12915 }, { "epoch": 4.01, "learning_rate": 4.836967393478696e-06, "loss": 0.0213, "step": 12916 }, { "epoch": 4.01, "learning_rate": 4.8365673134626925e-06, "loss": 0.0093, "step": 12917 }, { "epoch": 4.01, "learning_rate": 4.83616723344669e-06, "loss": 0.0154, "step": 12918 }, { "epoch": 4.01, "learning_rate": 4.835767153430687e-06, "loss": 0.0187, "step": 12919 }, { "epoch": 4.01, "learning_rate": 4.835367073414683e-06, "loss": 0.0736, "step": 12920 }, { "epoch": 4.01, "learning_rate": 4.83496699339868e-06, "loss": 0.0312, "step": 12921 }, { "epoch": 4.01, "learning_rate": 4.834566913382677e-06, "loss": 0.0257, "step": 12922 }, { "epoch": 4.01, "learning_rate": 4.8341668333666735e-06, "loss": 0.0109, "step": 12923 }, { "epoch": 4.01, "learning_rate": 4.83376675335067e-06, "loss": 0.0085, "step": 12924 }, { "epoch": 4.01, "learning_rate": 4.833366673334668e-06, "loss": 0.0808, "step": 12925 }, { "epoch": 4.01, "learning_rate": 4.832966593318664e-06, "loss": 0.0406, "step": 12926 }, { "epoch": 4.01, "learning_rate": 4.832566513302661e-06, "loss": 0.0242, "step": 12927 }, { "epoch": 4.01, "learning_rate": 4.832166433286658e-06, "loss": 0.021, "step": 12928 }, { "epoch": 4.01, "learning_rate": 4.831766353270654e-06, "loss": 0.0183, "step": 12929 }, { "epoch": 4.01, "learning_rate": 4.831366273254651e-06, "loss": 0.016, "step": 12930 }, { "epoch": 4.01, "learning_rate": 4.830966193238648e-06, "loss": 0.0177, "step": 12931 }, { "epoch": 4.01, "learning_rate": 4.830566113222645e-06, "loss": 0.1107, "step": 12932 }, { "epoch": 4.01, "learning_rate": 4.830166033206642e-06, "loss": 0.0268, "step": 12933 }, { "epoch": 4.01, "learning_rate": 4.829765953190639e-06, "loss": 0.0077, "step": 12934 }, { "epoch": 4.01, "learning_rate": 4.829365873174635e-06, "loss": 0.0141, "step": 12935 }, { "epoch": 4.02, "learning_rate": 4.828965793158632e-06, "loss": 0.0094, "step": 12936 }, { "epoch": 4.02, "learning_rate": 4.828565713142629e-06, "loss": 0.0112, "step": 12937 }, { "epoch": 4.02, "learning_rate": 4.8281656331266254e-06, "loss": 0.0176, "step": 12938 }, { "epoch": 4.02, "learning_rate": 4.827765553110623e-06, "loss": 0.0248, "step": 12939 }, { "epoch": 4.02, "learning_rate": 4.82736547309462e-06, "loss": 0.0408, "step": 12940 }, { "epoch": 4.02, "learning_rate": 4.8269653930786155e-06, "loss": 0.0109, "step": 12941 }, { "epoch": 4.02, "learning_rate": 4.826565313062613e-06, "loss": 0.031, "step": 12942 }, { "epoch": 4.02, "learning_rate": 4.82616523304661e-06, "loss": 0.0286, "step": 12943 }, { "epoch": 4.02, "learning_rate": 4.825765153030606e-06, "loss": 0.0162, "step": 12944 }, { "epoch": 4.02, "learning_rate": 4.825365073014603e-06, "loss": 0.0163, "step": 12945 }, { "epoch": 4.02, "learning_rate": 4.824964992998601e-06, "loss": 0.0268, "step": 12946 }, { "epoch": 4.02, "learning_rate": 4.8245649129825965e-06, "loss": 0.0139, "step": 12947 }, { "epoch": 4.02, "learning_rate": 4.824164832966593e-06, "loss": 0.0896, "step": 12948 }, { "epoch": 4.02, "learning_rate": 4.823764752950591e-06, "loss": 0.0125, "step": 12949 }, { "epoch": 4.02, "learning_rate": 4.823364672934587e-06, "loss": 0.0211, "step": 12950 }, { "epoch": 4.02, "learning_rate": 4.822964592918584e-06, "loss": 0.0097, "step": 12951 }, { "epoch": 4.02, "learning_rate": 4.822564512902581e-06, "loss": 0.0144, "step": 12952 }, { "epoch": 4.02, "learning_rate": 4.822164432886578e-06, "loss": 0.0217, "step": 12953 }, { "epoch": 4.02, "learning_rate": 4.821764352870574e-06, "loss": 0.1095, "step": 12954 }, { "epoch": 4.02, "learning_rate": 4.821364272854571e-06, "loss": 0.0179, "step": 12955 }, { "epoch": 4.02, "learning_rate": 4.820964192838568e-06, "loss": 0.0118, "step": 12956 }, { "epoch": 4.02, "learning_rate": 4.820564112822565e-06, "loss": 0.0308, "step": 12957 }, { "epoch": 4.02, "learning_rate": 4.820164032806562e-06, "loss": 0.0129, "step": 12958 }, { "epoch": 4.02, "learning_rate": 4.819763952790558e-06, "loss": 0.0374, "step": 12959 }, { "epoch": 4.02, "learning_rate": 4.819363872774555e-06, "loss": 0.0152, "step": 12960 }, { "epoch": 4.02, "learning_rate": 4.818963792758552e-06, "loss": 0.0129, "step": 12961 }, { "epoch": 4.02, "learning_rate": 4.8185637127425485e-06, "loss": 0.0161, "step": 12962 }, { "epoch": 4.02, "learning_rate": 4.818163632726546e-06, "loss": 0.0299, "step": 12963 }, { "epoch": 4.02, "learning_rate": 4.817763552710543e-06, "loss": 0.0196, "step": 12964 }, { "epoch": 4.02, "learning_rate": 4.817363472694539e-06, "loss": 0.1068, "step": 12965 }, { "epoch": 4.02, "learning_rate": 4.816963392678536e-06, "loss": 0.0252, "step": 12966 }, { "epoch": 4.02, "learning_rate": 4.816563312662533e-06, "loss": 0.0078, "step": 12967 }, { "epoch": 4.02, "learning_rate": 4.816163232646529e-06, "loss": 0.0114, "step": 12968 }, { "epoch": 4.03, "learning_rate": 4.815763152630526e-06, "loss": 0.0124, "step": 12969 }, { "epoch": 4.03, "learning_rate": 4.815363072614524e-06, "loss": 0.0077, "step": 12970 }, { "epoch": 4.03, "learning_rate": 4.81496299259852e-06, "loss": 0.0272, "step": 12971 }, { "epoch": 4.03, "learning_rate": 4.814562912582517e-06, "loss": 0.0259, "step": 12972 }, { "epoch": 4.03, "learning_rate": 4.814162832566514e-06, "loss": 0.0061, "step": 12973 }, { "epoch": 4.03, "learning_rate": 4.81376275255051e-06, "loss": 0.0136, "step": 12974 }, { "epoch": 4.03, "learning_rate": 4.813362672534507e-06, "loss": 0.0145, "step": 12975 }, { "epoch": 4.03, "learning_rate": 4.812962592518504e-06, "loss": 0.0134, "step": 12976 }, { "epoch": 4.03, "learning_rate": 4.812562512502501e-06, "loss": 0.0884, "step": 12977 }, { "epoch": 4.03, "learning_rate": 4.812162432486498e-06, "loss": 0.0273, "step": 12978 }, { "epoch": 4.03, "learning_rate": 4.811762352470495e-06, "loss": 0.009, "step": 12979 }, { "epoch": 4.03, "learning_rate": 4.811362272454491e-06, "loss": 0.0173, "step": 12980 }, { "epoch": 4.03, "learning_rate": 4.810962192438488e-06, "loss": 0.0963, "step": 12981 }, { "epoch": 4.03, "learning_rate": 4.810562112422485e-06, "loss": 0.0398, "step": 12982 }, { "epoch": 4.03, "learning_rate": 4.810162032406481e-06, "loss": 0.0212, "step": 12983 }, { "epoch": 4.03, "learning_rate": 4.809761952390479e-06, "loss": 0.0174, "step": 12984 }, { "epoch": 4.03, "learning_rate": 4.809361872374475e-06, "loss": 0.0183, "step": 12985 }, { "epoch": 4.03, "learning_rate": 4.808961792358472e-06, "loss": 0.0173, "step": 12986 }, { "epoch": 4.03, "learning_rate": 4.808561712342469e-06, "loss": 0.0143, "step": 12987 }, { "epoch": 4.03, "learning_rate": 4.808161632326466e-06, "loss": 0.0138, "step": 12988 }, { "epoch": 4.03, "learning_rate": 4.807761552310462e-06, "loss": 0.0136, "step": 12989 }, { "epoch": 4.03, "learning_rate": 4.80736147229446e-06, "loss": 0.0151, "step": 12990 }, { "epoch": 4.03, "learning_rate": 4.806961392278456e-06, "loss": 0.0456, "step": 12991 }, { "epoch": 4.03, "learning_rate": 4.8065613122624524e-06, "loss": 0.0306, "step": 12992 }, { "epoch": 4.03, "learning_rate": 4.80616123224645e-06, "loss": 0.0127, "step": 12993 }, { "epoch": 4.03, "learning_rate": 4.805761152230447e-06, "loss": 0.0245, "step": 12994 }, { "epoch": 4.03, "learning_rate": 4.805361072214443e-06, "loss": 0.0098, "step": 12995 }, { "epoch": 4.03, "learning_rate": 4.80496099219844e-06, "loss": 0.0342, "step": 12996 }, { "epoch": 4.03, "learning_rate": 4.8045609121824376e-06, "loss": 0.0134, "step": 12997 }, { "epoch": 4.03, "learning_rate": 4.804160832166433e-06, "loss": 0.0099, "step": 12998 }, { "epoch": 4.03, "learning_rate": 4.80376075215043e-06, "loss": 0.0168, "step": 12999 }, { "epoch": 4.03, "learning_rate": 4.803360672134428e-06, "loss": 0.009, "step": 13000 }, { "epoch": 4.04, "learning_rate": 4.802960592118424e-06, "loss": 0.0176, "step": 13001 }, { "epoch": 4.04, "learning_rate": 4.802560512102421e-06, "loss": 0.0084, "step": 13002 }, { "epoch": 4.04, "learning_rate": 4.802160432086418e-06, "loss": 0.0306, "step": 13003 }, { "epoch": 4.04, "learning_rate": 4.801760352070414e-06, "loss": 0.0998, "step": 13004 }, { "epoch": 4.04, "learning_rate": 4.801360272054411e-06, "loss": 0.0135, "step": 13005 }, { "epoch": 4.04, "learning_rate": 4.800960192038408e-06, "loss": 0.0177, "step": 13006 }, { "epoch": 4.04, "learning_rate": 4.800560112022405e-06, "loss": 0.025, "step": 13007 }, { "epoch": 4.04, "learning_rate": 4.800160032006402e-06, "loss": 0.0187, "step": 13008 }, { "epoch": 4.04, "learning_rate": 4.799759951990399e-06, "loss": 0.0148, "step": 13009 }, { "epoch": 4.04, "learning_rate": 4.799359871974395e-06, "loss": 0.0085, "step": 13010 }, { "epoch": 4.04, "learning_rate": 4.798959791958392e-06, "loss": 0.0209, "step": 13011 }, { "epoch": 4.04, "learning_rate": 4.798559711942389e-06, "loss": 0.0235, "step": 13012 }, { "epoch": 4.04, "learning_rate": 4.798159631926385e-06, "loss": 0.0207, "step": 13013 }, { "epoch": 4.04, "learning_rate": 4.797759551910383e-06, "loss": 0.0107, "step": 13014 }, { "epoch": 4.04, "learning_rate": 4.79735947189438e-06, "loss": 0.0252, "step": 13015 }, { "epoch": 4.04, "learning_rate": 4.796959391878376e-06, "loss": 0.0069, "step": 13016 }, { "epoch": 4.04, "learning_rate": 4.796559311862373e-06, "loss": 0.0093, "step": 13017 }, { "epoch": 4.04, "learning_rate": 4.79615923184637e-06, "loss": 0.0243, "step": 13018 }, { "epoch": 4.04, "learning_rate": 4.795759151830366e-06, "loss": 0.0256, "step": 13019 }, { "epoch": 4.04, "learning_rate": 4.795359071814363e-06, "loss": 0.0468, "step": 13020 }, { "epoch": 4.04, "learning_rate": 4.7949589917983606e-06, "loss": 0.0145, "step": 13021 }, { "epoch": 4.04, "learning_rate": 4.794558911782357e-06, "loss": 0.0299, "step": 13022 }, { "epoch": 4.04, "learning_rate": 4.794158831766353e-06, "loss": 0.0185, "step": 13023 }, { "epoch": 4.04, "learning_rate": 4.793758751750351e-06, "loss": 0.0154, "step": 13024 }, { "epoch": 4.04, "learning_rate": 4.793358671734347e-06, "loss": 0.103, "step": 13025 }, { "epoch": 4.04, "learning_rate": 4.792958591718344e-06, "loss": 0.0106, "step": 13026 }, { "epoch": 4.04, "learning_rate": 4.792558511702341e-06, "loss": 0.0099, "step": 13027 }, { "epoch": 4.04, "learning_rate": 4.792158431686338e-06, "loss": 0.0071, "step": 13028 }, { "epoch": 4.04, "learning_rate": 4.791758351670334e-06, "loss": 0.0163, "step": 13029 }, { "epoch": 4.04, "learning_rate": 4.791358271654331e-06, "loss": 0.0421, "step": 13030 }, { "epoch": 4.04, "learning_rate": 4.790958191638328e-06, "loss": 0.0165, "step": 13031 }, { "epoch": 4.04, "learning_rate": 4.790558111622325e-06, "loss": 0.0082, "step": 13032 }, { "epoch": 4.05, "learning_rate": 4.790158031606322e-06, "loss": 0.0111, "step": 13033 }, { "epoch": 4.05, "learning_rate": 4.789757951590318e-06, "loss": 0.0202, "step": 13034 }, { "epoch": 4.05, "learning_rate": 4.789357871574316e-06, "loss": 0.0139, "step": 13035 }, { "epoch": 4.05, "learning_rate": 4.788957791558312e-06, "loss": 0.0165, "step": 13036 }, { "epoch": 4.05, "learning_rate": 4.788557711542308e-06, "loss": 0.0137, "step": 13037 }, { "epoch": 4.05, "learning_rate": 4.788157631526306e-06, "loss": 0.0195, "step": 13038 }, { "epoch": 4.05, "learning_rate": 4.787757551510303e-06, "loss": 0.0067, "step": 13039 }, { "epoch": 4.05, "learning_rate": 4.787357471494299e-06, "loss": 0.0209, "step": 13040 }, { "epoch": 4.05, "learning_rate": 4.786957391478296e-06, "loss": 0.0068, "step": 13041 }, { "epoch": 4.05, "learning_rate": 4.786557311462293e-06, "loss": 0.0333, "step": 13042 }, { "epoch": 4.05, "learning_rate": 4.786157231446289e-06, "loss": 0.1456, "step": 13043 }, { "epoch": 4.05, "learning_rate": 4.785757151430286e-06, "loss": 0.0715, "step": 13044 }, { "epoch": 4.05, "learning_rate": 4.785357071414284e-06, "loss": 0.0154, "step": 13045 }, { "epoch": 4.05, "learning_rate": 4.78495699139828e-06, "loss": 0.0082, "step": 13046 }, { "epoch": 4.05, "learning_rate": 4.784556911382277e-06, "loss": 0.0247, "step": 13047 }, { "epoch": 4.05, "learning_rate": 4.784156831366274e-06, "loss": 0.1011, "step": 13048 }, { "epoch": 4.05, "learning_rate": 4.78375675135027e-06, "loss": 0.0236, "step": 13049 }, { "epoch": 4.05, "learning_rate": 4.783356671334267e-06, "loss": 0.0146, "step": 13050 }, { "epoch": 4.05, "learning_rate": 4.782956591318264e-06, "loss": 0.014, "step": 13051 }, { "epoch": 4.05, "learning_rate": 4.782556511302261e-06, "loss": 0.0573, "step": 13052 }, { "epoch": 4.05, "learning_rate": 4.782156431286258e-06, "loss": 0.0321, "step": 13053 }, { "epoch": 4.05, "learning_rate": 4.781756351270254e-06, "loss": 0.0818, "step": 13054 }, { "epoch": 4.05, "learning_rate": 4.781356271254251e-06, "loss": 0.0165, "step": 13055 }, { "epoch": 4.05, "learning_rate": 4.780956191238248e-06, "loss": 0.0119, "step": 13056 }, { "epoch": 4.05, "learning_rate": 4.780556111222245e-06, "loss": 0.0251, "step": 13057 }, { "epoch": 4.05, "learning_rate": 4.780156031206241e-06, "loss": 0.0146, "step": 13058 }, { "epoch": 4.05, "learning_rate": 4.779755951190239e-06, "loss": 0.0211, "step": 13059 }, { "epoch": 4.05, "learning_rate": 4.7793558711742356e-06, "loss": 0.0103, "step": 13060 }, { "epoch": 4.05, "learning_rate": 4.778955791158231e-06, "loss": 0.0222, "step": 13061 }, { "epoch": 4.05, "learning_rate": 4.778555711142229e-06, "loss": 0.022, "step": 13062 }, { "epoch": 4.05, "learning_rate": 4.778155631126226e-06, "loss": 0.0591, "step": 13063 }, { "epoch": 4.05, "learning_rate": 4.777755551110222e-06, "loss": 0.0199, "step": 13064 }, { "epoch": 4.06, "learning_rate": 4.777355471094219e-06, "loss": 0.0181, "step": 13065 }, { "epoch": 4.06, "learning_rate": 4.7769553910782165e-06, "loss": 0.037, "step": 13066 }, { "epoch": 4.06, "learning_rate": 4.776555311062212e-06, "loss": 0.0289, "step": 13067 }, { "epoch": 4.06, "learning_rate": 4.776155231046209e-06, "loss": 0.0295, "step": 13068 }, { "epoch": 4.06, "learning_rate": 4.775755151030207e-06, "loss": 0.0253, "step": 13069 }, { "epoch": 4.06, "learning_rate": 4.775355071014203e-06, "loss": 0.0064, "step": 13070 }, { "epoch": 4.06, "learning_rate": 4.7749549909982e-06, "loss": 0.0101, "step": 13071 }, { "epoch": 4.06, "learning_rate": 4.774554910982197e-06, "loss": 0.0258, "step": 13072 }, { "epoch": 4.06, "learning_rate": 4.774154830966193e-06, "loss": 0.0117, "step": 13073 }, { "epoch": 4.06, "learning_rate": 4.77375475095019e-06, "loss": 0.112, "step": 13074 }, { "epoch": 4.06, "learning_rate": 4.7733546709341876e-06, "loss": 0.0114, "step": 13075 }, { "epoch": 4.06, "learning_rate": 4.772954590918184e-06, "loss": 0.0156, "step": 13076 }, { "epoch": 4.06, "learning_rate": 4.772554510902181e-06, "loss": 0.0078, "step": 13077 }, { "epoch": 4.06, "learning_rate": 4.772154430886178e-06, "loss": 0.0211, "step": 13078 }, { "epoch": 4.06, "learning_rate": 4.771754350870174e-06, "loss": 0.0096, "step": 13079 }, { "epoch": 4.06, "learning_rate": 4.771354270854171e-06, "loss": 0.0078, "step": 13080 }, { "epoch": 4.06, "learning_rate": 4.770954190838168e-06, "loss": 0.0214, "step": 13081 }, { "epoch": 4.06, "learning_rate": 4.770554110822165e-06, "loss": 0.0153, "step": 13082 }, { "epoch": 4.06, "learning_rate": 4.770154030806162e-06, "loss": 0.0172, "step": 13083 }, { "epoch": 4.06, "learning_rate": 4.769753950790159e-06, "loss": 0.0142, "step": 13084 }, { "epoch": 4.06, "learning_rate": 4.769353870774155e-06, "loss": 0.0107, "step": 13085 }, { "epoch": 4.06, "learning_rate": 4.768953790758152e-06, "loss": 0.0263, "step": 13086 }, { "epoch": 4.06, "learning_rate": 4.768553710742149e-06, "loss": 0.0145, "step": 13087 }, { "epoch": 4.06, "learning_rate": 4.768153630726145e-06, "loss": 0.0196, "step": 13088 }, { "epoch": 4.06, "learning_rate": 4.767753550710143e-06, "loss": 0.0227, "step": 13089 }, { "epoch": 4.06, "learning_rate": 4.7673534706941395e-06, "loss": 0.0253, "step": 13090 }, { "epoch": 4.06, "learning_rate": 4.766953390678136e-06, "loss": 0.0174, "step": 13091 }, { "epoch": 4.06, "learning_rate": 4.766553310662133e-06, "loss": 0.0214, "step": 13092 }, { "epoch": 4.06, "learning_rate": 4.76615323064613e-06, "loss": 0.0191, "step": 13093 }, { "epoch": 4.06, "learning_rate": 4.765753150630126e-06, "loss": 0.0107, "step": 13094 }, { "epoch": 4.06, "learning_rate": 4.765353070614123e-06, "loss": 0.0113, "step": 13095 }, { "epoch": 4.06, "learning_rate": 4.7649529905981205e-06, "loss": 0.0118, "step": 13096 }, { "epoch": 4.07, "learning_rate": 4.764552910582117e-06, "loss": 0.0292, "step": 13097 }, { "epoch": 4.07, "learning_rate": 4.764152830566114e-06, "loss": 0.0049, "step": 13098 }, { "epoch": 4.07, "learning_rate": 4.7637527505501106e-06, "loss": 0.0071, "step": 13099 }, { "epoch": 4.07, "learning_rate": 4.763352670534107e-06, "loss": 0.0142, "step": 13100 }, { "epoch": 4.07, "learning_rate": 4.762952590518104e-06, "loss": 0.12, "step": 13101 }, { "epoch": 4.07, "learning_rate": 4.762552510502101e-06, "loss": 0.0225, "step": 13102 }, { "epoch": 4.07, "learning_rate": 4.762152430486098e-06, "loss": 0.0268, "step": 13103 }, { "epoch": 4.07, "learning_rate": 4.761752350470095e-06, "loss": 0.0069, "step": 13104 }, { "epoch": 4.07, "learning_rate": 4.761352270454091e-06, "loss": 0.006, "step": 13105 }, { "epoch": 4.07, "learning_rate": 4.760952190438088e-06, "loss": 0.0522, "step": 13106 }, { "epoch": 4.07, "learning_rate": 4.760552110422085e-06, "loss": 0.0078, "step": 13107 }, { "epoch": 4.07, "learning_rate": 4.760152030406082e-06, "loss": 0.0135, "step": 13108 }, { "epoch": 4.07, "learning_rate": 4.759751950390078e-06, "loss": 0.0074, "step": 13109 }, { "epoch": 4.07, "learning_rate": 4.759351870374076e-06, "loss": 0.0257, "step": 13110 }, { "epoch": 4.07, "learning_rate": 4.758951790358072e-06, "loss": 0.0169, "step": 13111 }, { "epoch": 4.07, "learning_rate": 4.758551710342068e-06, "loss": 0.0306, "step": 13112 }, { "epoch": 4.07, "learning_rate": 4.758151630326066e-06, "loss": 0.0398, "step": 13113 }, { "epoch": 4.07, "learning_rate": 4.7577515503100626e-06, "loss": 0.0119, "step": 13114 }, { "epoch": 4.07, "learning_rate": 4.757351470294059e-06, "loss": 0.0098, "step": 13115 }, { "epoch": 4.07, "learning_rate": 4.756951390278056e-06, "loss": 0.0132, "step": 13116 }, { "epoch": 4.07, "learning_rate": 4.756551310262053e-06, "loss": 0.0104, "step": 13117 }, { "epoch": 4.07, "learning_rate": 4.756151230246049e-06, "loss": 0.0236, "step": 13118 }, { "epoch": 4.07, "learning_rate": 4.755751150230046e-06, "loss": 0.0197, "step": 13119 }, { "epoch": 4.07, "learning_rate": 4.7553510702140435e-06, "loss": 0.0156, "step": 13120 }, { "epoch": 4.07, "learning_rate": 4.75495099019804e-06, "loss": 0.0115, "step": 13121 }, { "epoch": 4.07, "learning_rate": 4.754550910182037e-06, "loss": 0.0294, "step": 13122 }, { "epoch": 4.07, "learning_rate": 4.754150830166034e-06, "loss": 0.0866, "step": 13123 }, { "epoch": 4.07, "learning_rate": 4.75375075015003e-06, "loss": 0.0461, "step": 13124 }, { "epoch": 4.07, "learning_rate": 4.753350670134027e-06, "loss": 0.0312, "step": 13125 }, { "epoch": 4.07, "learning_rate": 4.752950590118024e-06, "loss": 0.0221, "step": 13126 }, { "epoch": 4.07, "learning_rate": 4.752550510102021e-06, "loss": 0.0234, "step": 13127 }, { "epoch": 4.07, "learning_rate": 4.752150430086018e-06, "loss": 0.0182, "step": 13128 }, { "epoch": 4.07, "learning_rate": 4.7517503500700145e-06, "loss": 0.0229, "step": 13129 }, { "epoch": 4.08, "learning_rate": 4.751350270054011e-06, "loss": 0.0141, "step": 13130 }, { "epoch": 4.08, "learning_rate": 4.750950190038008e-06, "loss": 0.0164, "step": 13131 }, { "epoch": 4.08, "learning_rate": 4.750550110022005e-06, "loss": 0.0106, "step": 13132 }, { "epoch": 4.08, "learning_rate": 4.750150030006001e-06, "loss": 0.0311, "step": 13133 }, { "epoch": 4.08, "learning_rate": 4.749749949989999e-06, "loss": 0.0314, "step": 13134 }, { "epoch": 4.08, "learning_rate": 4.7493498699739955e-06, "loss": 0.0162, "step": 13135 }, { "epoch": 4.08, "learning_rate": 4.748949789957991e-06, "loss": 0.0066, "step": 13136 }, { "epoch": 4.08, "learning_rate": 4.748549709941989e-06, "loss": 0.0181, "step": 13137 }, { "epoch": 4.08, "learning_rate": 4.7481496299259856e-06, "loss": 0.0233, "step": 13138 }, { "epoch": 4.08, "learning_rate": 4.747749549909982e-06, "loss": 0.0178, "step": 13139 }, { "epoch": 4.08, "learning_rate": 4.747349469893979e-06, "loss": 0.011, "step": 13140 }, { "epoch": 4.08, "learning_rate": 4.7469493898779765e-06, "loss": 0.0205, "step": 13141 }, { "epoch": 4.08, "learning_rate": 4.746549309861973e-06, "loss": 0.0157, "step": 13142 }, { "epoch": 4.08, "learning_rate": 4.746149229845969e-06, "loss": 0.0198, "step": 13143 }, { "epoch": 4.08, "learning_rate": 4.7457491498299665e-06, "loss": 0.0069, "step": 13144 }, { "epoch": 4.08, "learning_rate": 4.745349069813963e-06, "loss": 0.0058, "step": 13145 }, { "epoch": 4.08, "learning_rate": 4.74494898979796e-06, "loss": 0.0169, "step": 13146 }, { "epoch": 4.08, "learning_rate": 4.744548909781957e-06, "loss": 0.0153, "step": 13147 }, { "epoch": 4.08, "learning_rate": 4.744148829765954e-06, "loss": 0.0098, "step": 13148 }, { "epoch": 4.08, "learning_rate": 4.74374874974995e-06, "loss": 0.0103, "step": 13149 }, { "epoch": 4.08, "learning_rate": 4.743348669733947e-06, "loss": 0.1126, "step": 13150 }, { "epoch": 4.08, "learning_rate": 4.742948589717944e-06, "loss": 0.0177, "step": 13151 }, { "epoch": 4.08, "learning_rate": 4.742548509701941e-06, "loss": 0.011, "step": 13152 }, { "epoch": 4.08, "learning_rate": 4.7421484296859376e-06, "loss": 0.0137, "step": 13153 }, { "epoch": 4.08, "learning_rate": 4.741748349669934e-06, "loss": 0.0203, "step": 13154 }, { "epoch": 4.08, "learning_rate": 4.741348269653931e-06, "loss": 0.0182, "step": 13155 }, { "epoch": 4.08, "learning_rate": 4.740948189637928e-06, "loss": 0.0138, "step": 13156 }, { "epoch": 4.08, "learning_rate": 4.740548109621924e-06, "loss": 0.0114, "step": 13157 }, { "epoch": 4.08, "learning_rate": 4.740148029605922e-06, "loss": 0.0197, "step": 13158 }, { "epoch": 4.08, "learning_rate": 4.7397479495899185e-06, "loss": 0.072, "step": 13159 }, { "epoch": 4.08, "learning_rate": 4.739347869573915e-06, "loss": 0.0087, "step": 13160 }, { "epoch": 4.08, "learning_rate": 4.738947789557912e-06, "loss": 0.0129, "step": 13161 }, { "epoch": 4.09, "learning_rate": 4.738547709541909e-06, "loss": 0.0316, "step": 13162 }, { "epoch": 4.09, "learning_rate": 4.738147629525905e-06, "loss": 0.0223, "step": 13163 }, { "epoch": 4.09, "learning_rate": 4.737747549509902e-06, "loss": 0.0146, "step": 13164 }, { "epoch": 4.09, "learning_rate": 4.7373474694938995e-06, "loss": 0.0186, "step": 13165 }, { "epoch": 4.09, "learning_rate": 4.736947389477896e-06, "loss": 0.0265, "step": 13166 }, { "epoch": 4.09, "learning_rate": 4.736547309461893e-06, "loss": 0.0231, "step": 13167 }, { "epoch": 4.09, "learning_rate": 4.7361472294458895e-06, "loss": 0.0358, "step": 13168 }, { "epoch": 4.09, "learning_rate": 4.735747149429886e-06, "loss": 0.0203, "step": 13169 }, { "epoch": 4.09, "learning_rate": 4.735347069413883e-06, "loss": 0.015, "step": 13170 }, { "epoch": 4.09, "learning_rate": 4.73494698939788e-06, "loss": 0.0076, "step": 13171 }, { "epoch": 4.09, "learning_rate": 4.734546909381877e-06, "loss": 0.0278, "step": 13172 }, { "epoch": 4.09, "learning_rate": 4.734146829365874e-06, "loss": 0.0085, "step": 13173 }, { "epoch": 4.09, "learning_rate": 4.7337467493498705e-06, "loss": 0.0223, "step": 13174 }, { "epoch": 4.09, "learning_rate": 4.733346669333867e-06, "loss": 0.0849, "step": 13175 }, { "epoch": 4.09, "learning_rate": 4.732946589317864e-06, "loss": 0.013, "step": 13176 }, { "epoch": 4.09, "learning_rate": 4.7325465093018606e-06, "loss": 0.008, "step": 13177 }, { "epoch": 4.09, "learning_rate": 4.732146429285857e-06, "loss": 0.0158, "step": 13178 }, { "epoch": 4.09, "learning_rate": 4.731746349269855e-06, "loss": 0.0171, "step": 13179 }, { "epoch": 4.09, "learning_rate": 4.7313462692538515e-06, "loss": 0.0091, "step": 13180 }, { "epoch": 4.09, "learning_rate": 4.730946189237848e-06, "loss": 0.0109, "step": 13181 }, { "epoch": 4.09, "learning_rate": 4.730546109221845e-06, "loss": 0.0121, "step": 13182 }, { "epoch": 4.09, "learning_rate": 4.7301460292058415e-06, "loss": 0.0137, "step": 13183 }, { "epoch": 4.09, "learning_rate": 4.729745949189838e-06, "loss": 0.0301, "step": 13184 }, { "epoch": 4.09, "learning_rate": 4.729345869173835e-06, "loss": 0.0064, "step": 13185 }, { "epoch": 4.09, "learning_rate": 4.7289457891578324e-06, "loss": 0.0244, "step": 13186 }, { "epoch": 4.09, "learning_rate": 4.728545709141828e-06, "loss": 0.0758, "step": 13187 }, { "epoch": 4.09, "learning_rate": 4.728145629125826e-06, "loss": 0.0184, "step": 13188 }, { "epoch": 4.09, "learning_rate": 4.7277455491098225e-06, "loss": 0.0145, "step": 13189 }, { "epoch": 4.09, "learning_rate": 4.727345469093819e-06, "loss": 0.0443, "step": 13190 }, { "epoch": 4.09, "learning_rate": 4.726945389077816e-06, "loss": 0.0168, "step": 13191 }, { "epoch": 4.09, "learning_rate": 4.7265453090618126e-06, "loss": 0.0118, "step": 13192 }, { "epoch": 4.09, "learning_rate": 4.726145229045809e-06, "loss": 0.035, "step": 13193 }, { "epoch": 4.1, "learning_rate": 4.725745149029806e-06, "loss": 0.0285, "step": 13194 }, { "epoch": 4.1, "learning_rate": 4.7253450690138035e-06, "loss": 0.014, "step": 13195 }, { "epoch": 4.1, "learning_rate": 4.7249449889978e-06, "loss": 0.0116, "step": 13196 }, { "epoch": 4.1, "learning_rate": 4.724544908981797e-06, "loss": 0.0309, "step": 13197 }, { "epoch": 4.1, "learning_rate": 4.7241448289657935e-06, "loss": 0.0223, "step": 13198 }, { "epoch": 4.1, "learning_rate": 4.72374474894979e-06, "loss": 0.0406, "step": 13199 }, { "epoch": 4.1, "learning_rate": 4.723344668933787e-06, "loss": 0.0308, "step": 13200 }, { "epoch": 4.1, "learning_rate": 4.722944588917784e-06, "loss": 0.0228, "step": 13201 }, { "epoch": 4.1, "learning_rate": 4.722544508901781e-06, "loss": 0.0314, "step": 13202 }, { "epoch": 4.1, "learning_rate": 4.722144428885778e-06, "loss": 0.0447, "step": 13203 }, { "epoch": 4.1, "learning_rate": 4.7217443488697745e-06, "loss": 0.0242, "step": 13204 }, { "epoch": 4.1, "learning_rate": 4.721344268853771e-06, "loss": 0.0243, "step": 13205 }, { "epoch": 4.1, "learning_rate": 4.720944188837768e-06, "loss": 0.0304, "step": 13206 }, { "epoch": 4.1, "learning_rate": 4.7205441088217645e-06, "loss": 0.0404, "step": 13207 }, { "epoch": 4.1, "learning_rate": 4.720144028805761e-06, "loss": 0.0241, "step": 13208 }, { "epoch": 4.1, "learning_rate": 4.719743948789759e-06, "loss": 0.0089, "step": 13209 }, { "epoch": 4.1, "learning_rate": 4.7193438687737555e-06, "loss": 0.013, "step": 13210 }, { "epoch": 4.1, "learning_rate": 4.718943788757752e-06, "loss": 0.0168, "step": 13211 }, { "epoch": 4.1, "learning_rate": 4.718543708741749e-06, "loss": 0.0157, "step": 13212 }, { "epoch": 4.1, "learning_rate": 4.7181436287257455e-06, "loss": 0.0843, "step": 13213 }, { "epoch": 4.1, "learning_rate": 4.717743548709742e-06, "loss": 0.0141, "step": 13214 }, { "epoch": 4.1, "learning_rate": 4.717343468693739e-06, "loss": 0.0527, "step": 13215 }, { "epoch": 4.1, "learning_rate": 4.716943388677736e-06, "loss": 0.04, "step": 13216 }, { "epoch": 4.1, "learning_rate": 4.716543308661733e-06, "loss": 0.0475, "step": 13217 }, { "epoch": 4.1, "learning_rate": 4.716143228645729e-06, "loss": 0.0105, "step": 13218 }, { "epoch": 4.1, "learning_rate": 4.7157431486297265e-06, "loss": 0.0281, "step": 13219 }, { "epoch": 4.1, "learning_rate": 4.715343068613723e-06, "loss": 0.0171, "step": 13220 }, { "epoch": 4.1, "learning_rate": 4.71494298859772e-06, "loss": 0.0341, "step": 13221 }, { "epoch": 4.1, "learning_rate": 4.7145429085817165e-06, "loss": 0.0324, "step": 13222 }, { "epoch": 4.1, "learning_rate": 4.714142828565714e-06, "loss": 0.011, "step": 13223 }, { "epoch": 4.1, "learning_rate": 4.713742748549711e-06, "loss": 0.0195, "step": 13224 }, { "epoch": 4.1, "learning_rate": 4.713342668533707e-06, "loss": 0.0517, "step": 13225 }, { "epoch": 4.11, "learning_rate": 4.712942588517704e-06, "loss": 0.0152, "step": 13226 }, { "epoch": 4.11, "learning_rate": 4.712542508501701e-06, "loss": 0.016, "step": 13227 }, { "epoch": 4.11, "learning_rate": 4.7121424284856975e-06, "loss": 0.0262, "step": 13228 }, { "epoch": 4.11, "learning_rate": 4.711742348469694e-06, "loss": 0.0297, "step": 13229 }, { "epoch": 4.11, "learning_rate": 4.711342268453692e-06, "loss": 0.0149, "step": 13230 }, { "epoch": 4.11, "learning_rate": 4.7109421884376876e-06, "loss": 0.0201, "step": 13231 }, { "epoch": 4.11, "learning_rate": 4.710542108421684e-06, "loss": 0.0123, "step": 13232 }, { "epoch": 4.11, "learning_rate": 4.710142028405682e-06, "loss": 0.0127, "step": 13233 }, { "epoch": 4.11, "learning_rate": 4.7097419483896785e-06, "loss": 0.0195, "step": 13234 }, { "epoch": 4.11, "learning_rate": 4.709341868373675e-06, "loss": 0.0089, "step": 13235 }, { "epoch": 4.11, "learning_rate": 4.708941788357672e-06, "loss": 0.0087, "step": 13236 }, { "epoch": 4.11, "learning_rate": 4.7085417083416685e-06, "loss": 0.019, "step": 13237 }, { "epoch": 4.11, "learning_rate": 4.708141628325665e-06, "loss": 0.0144, "step": 13238 }, { "epoch": 4.11, "learning_rate": 4.707741548309662e-06, "loss": 0.1065, "step": 13239 }, { "epoch": 4.11, "learning_rate": 4.7073414682936594e-06, "loss": 0.0219, "step": 13240 }, { "epoch": 4.11, "learning_rate": 4.706941388277656e-06, "loss": 0.0159, "step": 13241 }, { "epoch": 4.11, "learning_rate": 4.706541308261653e-06, "loss": 0.0328, "step": 13242 }, { "epoch": 4.11, "learning_rate": 4.7061412282456495e-06, "loss": 0.01, "step": 13243 }, { "epoch": 4.11, "learning_rate": 4.705741148229646e-06, "loss": 0.0085, "step": 13244 }, { "epoch": 4.11, "learning_rate": 4.705341068213643e-06, "loss": 0.0102, "step": 13245 }, { "epoch": 4.11, "learning_rate": 4.7049409881976395e-06, "loss": 0.0097, "step": 13246 }, { "epoch": 4.11, "learning_rate": 4.704540908181637e-06, "loss": 0.0324, "step": 13247 }, { "epoch": 4.11, "learning_rate": 4.704140828165634e-06, "loss": 0.0135, "step": 13248 }, { "epoch": 4.11, "learning_rate": 4.7037407481496305e-06, "loss": 0.0144, "step": 13249 }, { "epoch": 4.11, "learning_rate": 4.703340668133627e-06, "loss": 0.0139, "step": 13250 }, { "epoch": 4.11, "learning_rate": 4.702940588117624e-06, "loss": 0.0045, "step": 13251 }, { "epoch": 4.11, "learning_rate": 4.7025405081016205e-06, "loss": 0.0219, "step": 13252 }, { "epoch": 4.11, "learning_rate": 4.702140428085617e-06, "loss": 0.018, "step": 13253 }, { "epoch": 4.11, "learning_rate": 4.701740348069615e-06, "loss": 0.0193, "step": 13254 }, { "epoch": 4.11, "learning_rate": 4.701340268053611e-06, "loss": 0.0199, "step": 13255 }, { "epoch": 4.11, "learning_rate": 4.700940188037607e-06, "loss": 0.065, "step": 13256 }, { "epoch": 4.11, "learning_rate": 4.700540108021605e-06, "loss": 0.0111, "step": 13257 }, { "epoch": 4.11, "learning_rate": 4.7001400280056015e-06, "loss": 0.0202, "step": 13258 }, { "epoch": 4.12, "learning_rate": 4.699739947989598e-06, "loss": 0.0164, "step": 13259 }, { "epoch": 4.12, "learning_rate": 4.699339867973595e-06, "loss": 0.0082, "step": 13260 }, { "epoch": 4.12, "learning_rate": 4.698939787957592e-06, "loss": 0.0151, "step": 13261 }, { "epoch": 4.12, "learning_rate": 4.698539707941589e-06, "loss": 0.0195, "step": 13262 }, { "epoch": 4.12, "learning_rate": 4.698139627925585e-06, "loss": 0.0057, "step": 13263 }, { "epoch": 4.12, "learning_rate": 4.6977395479095824e-06, "loss": 0.0131, "step": 13264 }, { "epoch": 4.12, "learning_rate": 4.697339467893579e-06, "loss": 0.0139, "step": 13265 }, { "epoch": 4.12, "learning_rate": 4.696939387877576e-06, "loss": 0.0272, "step": 13266 }, { "epoch": 4.12, "learning_rate": 4.6965393078615725e-06, "loss": 0.0237, "step": 13267 }, { "epoch": 4.12, "learning_rate": 4.69613922784557e-06, "loss": 0.0324, "step": 13268 }, { "epoch": 4.12, "learning_rate": 4.695739147829566e-06, "loss": 0.0359, "step": 13269 }, { "epoch": 4.12, "learning_rate": 4.6953390678135626e-06, "loss": 0.0048, "step": 13270 }, { "epoch": 4.12, "learning_rate": 4.69493898779756e-06, "loss": 0.0143, "step": 13271 }, { "epoch": 4.12, "learning_rate": 4.694538907781557e-06, "loss": 0.0227, "step": 13272 }, { "epoch": 4.12, "learning_rate": 4.6941388277655535e-06, "loss": 0.0086, "step": 13273 }, { "epoch": 4.12, "learning_rate": 4.69373874774955e-06, "loss": 0.007, "step": 13274 }, { "epoch": 4.12, "learning_rate": 4.693338667733547e-06, "loss": 0.0129, "step": 13275 }, { "epoch": 4.12, "learning_rate": 4.6929385877175435e-06, "loss": 0.0209, "step": 13276 }, { "epoch": 4.12, "learning_rate": 4.69253850770154e-06, "loss": 0.0172, "step": 13277 }, { "epoch": 4.12, "learning_rate": 4.692138427685538e-06, "loss": 0.0257, "step": 13278 }, { "epoch": 4.12, "learning_rate": 4.6917383476695344e-06, "loss": 0.04, "step": 13279 }, { "epoch": 4.12, "learning_rate": 4.691338267653531e-06, "loss": 0.0153, "step": 13280 }, { "epoch": 4.12, "learning_rate": 4.690938187637528e-06, "loss": 0.0067, "step": 13281 }, { "epoch": 4.12, "learning_rate": 4.6905381076215245e-06, "loss": 0.0456, "step": 13282 }, { "epoch": 4.12, "learning_rate": 4.690138027605521e-06, "loss": 0.1346, "step": 13283 }, { "epoch": 4.12, "learning_rate": 4.689737947589518e-06, "loss": 0.0263, "step": 13284 }, { "epoch": 4.12, "learning_rate": 4.689337867573515e-06, "loss": 0.0527, "step": 13285 }, { "epoch": 4.12, "learning_rate": 4.688937787557512e-06, "loss": 0.0085, "step": 13286 }, { "epoch": 4.12, "learning_rate": 4.688537707541509e-06, "loss": 0.0055, "step": 13287 }, { "epoch": 4.12, "learning_rate": 4.6881376275255055e-06, "loss": 0.0149, "step": 13288 }, { "epoch": 4.12, "learning_rate": 4.687737547509502e-06, "loss": 0.0072, "step": 13289 }, { "epoch": 4.12, "learning_rate": 4.687337467493499e-06, "loss": 0.018, "step": 13290 }, { "epoch": 4.13, "learning_rate": 4.686937387477496e-06, "loss": 0.0325, "step": 13291 }, { "epoch": 4.13, "learning_rate": 4.686537307461493e-06, "loss": 0.0131, "step": 13292 }, { "epoch": 4.13, "learning_rate": 4.68613722744549e-06, "loss": 0.0129, "step": 13293 }, { "epoch": 4.13, "learning_rate": 4.685737147429486e-06, "loss": 0.0152, "step": 13294 }, { "epoch": 4.13, "learning_rate": 4.685337067413483e-06, "loss": 0.0155, "step": 13295 }, { "epoch": 4.13, "learning_rate": 4.68493698739748e-06, "loss": 0.0131, "step": 13296 }, { "epoch": 4.13, "learning_rate": 4.6845369073814765e-06, "loss": 0.0124, "step": 13297 }, { "epoch": 4.13, "learning_rate": 4.684136827365474e-06, "loss": 0.0142, "step": 13298 }, { "epoch": 4.13, "learning_rate": 4.683736747349471e-06, "loss": 0.0146, "step": 13299 }, { "epoch": 4.13, "learning_rate": 4.6833366673334665e-06, "loss": 0.0167, "step": 13300 }, { "epoch": 4.13, "learning_rate": 4.682936587317464e-06, "loss": 0.0257, "step": 13301 }, { "epoch": 4.13, "learning_rate": 4.682536507301461e-06, "loss": 0.0055, "step": 13302 }, { "epoch": 4.13, "learning_rate": 4.6821364272854574e-06, "loss": 0.0247, "step": 13303 }, { "epoch": 4.13, "learning_rate": 4.681736347269454e-06, "loss": 0.0563, "step": 13304 }, { "epoch": 4.13, "learning_rate": 4.681336267253452e-06, "loss": 0.0269, "step": 13305 }, { "epoch": 4.13, "learning_rate": 4.680936187237448e-06, "loss": 0.0631, "step": 13306 }, { "epoch": 4.13, "learning_rate": 4.680536107221444e-06, "loss": 0.0635, "step": 13307 }, { "epoch": 4.13, "learning_rate": 4.680136027205442e-06, "loss": 0.0321, "step": 13308 }, { "epoch": 4.13, "learning_rate": 4.679735947189438e-06, "loss": 0.0153, "step": 13309 }, { "epoch": 4.13, "learning_rate": 4.679335867173435e-06, "loss": 0.0199, "step": 13310 }, { "epoch": 4.13, "learning_rate": 4.678935787157432e-06, "loss": 0.0547, "step": 13311 }, { "epoch": 4.13, "learning_rate": 4.678535707141429e-06, "loss": 0.0198, "step": 13312 }, { "epoch": 4.13, "learning_rate": 4.678135627125425e-06, "loss": 0.0129, "step": 13313 }, { "epoch": 4.13, "learning_rate": 4.677735547109422e-06, "loss": 0.0088, "step": 13314 }, { "epoch": 4.13, "learning_rate": 4.677335467093419e-06, "loss": 0.0247, "step": 13315 }, { "epoch": 4.13, "learning_rate": 4.676935387077416e-06, "loss": 0.0327, "step": 13316 }, { "epoch": 4.13, "learning_rate": 4.676535307061413e-06, "loss": 0.0196, "step": 13317 }, { "epoch": 4.13, "learning_rate": 4.6761352270454094e-06, "loss": 0.0182, "step": 13318 }, { "epoch": 4.13, "learning_rate": 4.675735147029406e-06, "loss": 0.0158, "step": 13319 }, { "epoch": 4.13, "learning_rate": 4.675335067013403e-06, "loss": 0.0141, "step": 13320 }, { "epoch": 4.13, "learning_rate": 4.6749349869973995e-06, "loss": 0.017, "step": 13321 }, { "epoch": 4.13, "learning_rate": 4.674534906981397e-06, "loss": 0.0143, "step": 13322 }, { "epoch": 4.14, "learning_rate": 4.674134826965394e-06, "loss": 0.0474, "step": 13323 }, { "epoch": 4.14, "learning_rate": 4.67373474694939e-06, "loss": 0.055, "step": 13324 }, { "epoch": 4.14, "learning_rate": 4.673334666933387e-06, "loss": 0.0167, "step": 13325 }, { "epoch": 4.14, "learning_rate": 4.672934586917384e-06, "loss": 0.0346, "step": 13326 }, { "epoch": 4.14, "learning_rate": 4.6725345069013805e-06, "loss": 0.0119, "step": 13327 }, { "epoch": 4.14, "learning_rate": 4.672134426885377e-06, "loss": 0.0099, "step": 13328 }, { "epoch": 4.14, "learning_rate": 4.671734346869375e-06, "loss": 0.0088, "step": 13329 }, { "epoch": 4.14, "learning_rate": 4.671334266853371e-06, "loss": 0.0135, "step": 13330 }, { "epoch": 4.14, "learning_rate": 4.670934186837368e-06, "loss": 0.0291, "step": 13331 }, { "epoch": 4.14, "learning_rate": 4.670534106821365e-06, "loss": 0.0175, "step": 13332 }, { "epoch": 4.14, "learning_rate": 4.670134026805361e-06, "loss": 0.0379, "step": 13333 }, { "epoch": 4.14, "learning_rate": 4.669733946789358e-06, "loss": 0.0078, "step": 13334 }, { "epoch": 4.14, "learning_rate": 4.669333866773355e-06, "loss": 0.0189, "step": 13335 }, { "epoch": 4.14, "learning_rate": 4.668933786757352e-06, "loss": 0.0703, "step": 13336 }, { "epoch": 4.14, "learning_rate": 4.668533706741349e-06, "loss": 0.0284, "step": 13337 }, { "epoch": 4.14, "learning_rate": 4.668133626725345e-06, "loss": 0.0268, "step": 13338 }, { "epoch": 4.14, "learning_rate": 4.667733546709342e-06, "loss": 0.0219, "step": 13339 }, { "epoch": 4.14, "learning_rate": 4.667333466693339e-06, "loss": 0.0063, "step": 13340 }, { "epoch": 4.14, "learning_rate": 4.666933386677336e-06, "loss": 0.0223, "step": 13341 }, { "epoch": 4.14, "learning_rate": 4.6665333066613324e-06, "loss": 0.0145, "step": 13342 }, { "epoch": 4.14, "learning_rate": 4.66613322664533e-06, "loss": 0.0079, "step": 13343 }, { "epoch": 4.14, "learning_rate": 4.665733146629326e-06, "loss": 0.0089, "step": 13344 }, { "epoch": 4.14, "learning_rate": 4.6653330666133225e-06, "loss": 0.0583, "step": 13345 }, { "epoch": 4.14, "learning_rate": 4.66493298659732e-06, "loss": 0.0119, "step": 13346 }, { "epoch": 4.14, "learning_rate": 4.664532906581317e-06, "loss": 0.0097, "step": 13347 }, { "epoch": 4.14, "learning_rate": 4.664132826565313e-06, "loss": 0.0075, "step": 13348 }, { "epoch": 4.14, "learning_rate": 4.66373274654931e-06, "loss": 0.0546, "step": 13349 }, { "epoch": 4.14, "learning_rate": 4.663332666533308e-06, "loss": 0.0364, "step": 13350 }, { "epoch": 4.14, "learning_rate": 4.6629325865173035e-06, "loss": 0.0295, "step": 13351 }, { "epoch": 4.14, "learning_rate": 4.6625325065013e-06, "loss": 0.0076, "step": 13352 }, { "epoch": 4.14, "learning_rate": 4.662132426485298e-06, "loss": 0.0133, "step": 13353 }, { "epoch": 4.14, "learning_rate": 4.661732346469294e-06, "loss": 0.0272, "step": 13354 }, { "epoch": 4.15, "learning_rate": 4.661332266453291e-06, "loss": 0.0116, "step": 13355 }, { "epoch": 4.15, "learning_rate": 4.660932186437288e-06, "loss": 0.0654, "step": 13356 }, { "epoch": 4.15, "learning_rate": 4.6605321064212844e-06, "loss": 0.0066, "step": 13357 }, { "epoch": 4.15, "learning_rate": 4.660132026405281e-06, "loss": 0.0949, "step": 13358 }, { "epoch": 4.15, "learning_rate": 4.659731946389278e-06, "loss": 0.0157, "step": 13359 }, { "epoch": 4.15, "learning_rate": 4.659331866373275e-06, "loss": 0.0063, "step": 13360 }, { "epoch": 4.15, "learning_rate": 4.658931786357272e-06, "loss": 0.0085, "step": 13361 }, { "epoch": 4.15, "learning_rate": 4.658531706341269e-06, "loss": 0.089, "step": 13362 }, { "epoch": 4.15, "learning_rate": 4.658131626325265e-06, "loss": 0.0168, "step": 13363 }, { "epoch": 4.15, "learning_rate": 4.657731546309262e-06, "loss": 0.0056, "step": 13364 }, { "epoch": 4.15, "learning_rate": 4.657331466293259e-06, "loss": 0.0141, "step": 13365 }, { "epoch": 4.15, "learning_rate": 4.6569313862772555e-06, "loss": 0.0238, "step": 13366 }, { "epoch": 4.15, "learning_rate": 4.656531306261253e-06, "loss": 0.0144, "step": 13367 }, { "epoch": 4.15, "learning_rate": 4.65613122624525e-06, "loss": 0.0066, "step": 13368 }, { "epoch": 4.15, "learning_rate": 4.655731146229246e-06, "loss": 0.0054, "step": 13369 }, { "epoch": 4.15, "learning_rate": 4.655331066213243e-06, "loss": 0.0174, "step": 13370 }, { "epoch": 4.15, "learning_rate": 4.65493098619724e-06, "loss": 0.0062, "step": 13371 }, { "epoch": 4.15, "learning_rate": 4.654530906181236e-06, "loss": 0.0195, "step": 13372 }, { "epoch": 4.15, "learning_rate": 4.654130826165233e-06, "loss": 0.0101, "step": 13373 }, { "epoch": 4.15, "learning_rate": 4.653730746149231e-06, "loss": 0.0191, "step": 13374 }, { "epoch": 4.15, "learning_rate": 4.653330666133227e-06, "loss": 0.0378, "step": 13375 }, { "epoch": 4.15, "learning_rate": 4.652930586117224e-06, "loss": 0.0162, "step": 13376 }, { "epoch": 4.15, "learning_rate": 4.652530506101221e-06, "loss": 0.0061, "step": 13377 }, { "epoch": 4.15, "learning_rate": 4.652130426085217e-06, "loss": 0.0247, "step": 13378 }, { "epoch": 4.15, "learning_rate": 4.651730346069214e-06, "loss": 0.0241, "step": 13379 }, { "epoch": 4.15, "learning_rate": 4.651330266053211e-06, "loss": 0.0368, "step": 13380 }, { "epoch": 4.15, "learning_rate": 4.650930186037208e-06, "loss": 0.0364, "step": 13381 }, { "epoch": 4.15, "learning_rate": 4.650530106021204e-06, "loss": 0.0147, "step": 13382 }, { "epoch": 4.15, "learning_rate": 4.650130026005202e-06, "loss": 0.0224, "step": 13383 }, { "epoch": 4.15, "learning_rate": 4.649729945989198e-06, "loss": 0.0061, "step": 13384 }, { "epoch": 4.15, "learning_rate": 4.649329865973195e-06, "loss": 0.021, "step": 13385 }, { "epoch": 4.15, "learning_rate": 4.648929785957192e-06, "loss": 0.0162, "step": 13386 }, { "epoch": 4.16, "learning_rate": 4.648529705941188e-06, "loss": 0.019, "step": 13387 }, { "epoch": 4.16, "learning_rate": 4.648129625925186e-06, "loss": 0.0157, "step": 13388 }, { "epoch": 4.16, "learning_rate": 4.647729545909182e-06, "loss": 0.0276, "step": 13389 }, { "epoch": 4.16, "learning_rate": 4.647329465893179e-06, "loss": 0.0139, "step": 13390 }, { "epoch": 4.16, "learning_rate": 4.646929385877176e-06, "loss": 0.0079, "step": 13391 }, { "epoch": 4.16, "learning_rate": 4.646529305861173e-06, "loss": 0.0257, "step": 13392 }, { "epoch": 4.16, "learning_rate": 4.646129225845169e-06, "loss": 0.0071, "step": 13393 }, { "epoch": 4.16, "learning_rate": 4.645729145829166e-06, "loss": 0.0101, "step": 13394 }, { "epoch": 4.16, "learning_rate": 4.645329065813163e-06, "loss": 0.0787, "step": 13395 }, { "epoch": 4.16, "learning_rate": 4.6449289857971594e-06, "loss": 0.0215, "step": 13396 }, { "epoch": 4.16, "learning_rate": 4.644528905781157e-06, "loss": 0.0291, "step": 13397 }, { "epoch": 4.16, "learning_rate": 4.644128825765154e-06, "loss": 0.0376, "step": 13398 }, { "epoch": 4.16, "learning_rate": 4.64372874574915e-06, "loss": 0.0164, "step": 13399 }, { "epoch": 4.16, "learning_rate": 4.643328665733147e-06, "loss": 0.0863, "step": 13400 }, { "epoch": 4.16, "learning_rate": 4.642928585717144e-06, "loss": 0.0296, "step": 13401 }, { "epoch": 4.16, "learning_rate": 4.64252850570114e-06, "loss": 0.021, "step": 13402 }, { "epoch": 4.16, "learning_rate": 4.642128425685137e-06, "loss": 0.0064, "step": 13403 }, { "epoch": 4.16, "learning_rate": 4.641728345669135e-06, "loss": 0.011, "step": 13404 }, { "epoch": 4.16, "learning_rate": 4.641328265653131e-06, "loss": 0.0073, "step": 13405 }, { "epoch": 4.16, "learning_rate": 4.640928185637128e-06, "loss": 0.012, "step": 13406 }, { "epoch": 4.16, "learning_rate": 4.640528105621125e-06, "loss": 0.0254, "step": 13407 }, { "epoch": 4.16, "learning_rate": 4.640128025605121e-06, "loss": 0.0258, "step": 13408 }, { "epoch": 4.16, "learning_rate": 4.639727945589118e-06, "loss": 0.0203, "step": 13409 }, { "epoch": 4.16, "learning_rate": 4.639327865573115e-06, "loss": 0.036, "step": 13410 }, { "epoch": 4.16, "learning_rate": 4.638927785557112e-06, "loss": 0.0506, "step": 13411 }, { "epoch": 4.16, "learning_rate": 4.638527705541109e-06, "loss": 0.0317, "step": 13412 }, { "epoch": 4.16, "learning_rate": 4.638127625525106e-06, "loss": 0.0188, "step": 13413 }, { "epoch": 4.16, "learning_rate": 4.637727545509102e-06, "loss": 0.0212, "step": 13414 }, { "epoch": 4.16, "learning_rate": 4.637327465493099e-06, "loss": 0.0125, "step": 13415 }, { "epoch": 4.16, "learning_rate": 4.636927385477096e-06, "loss": 0.0519, "step": 13416 }, { "epoch": 4.16, "learning_rate": 4.636527305461092e-06, "loss": 0.008, "step": 13417 }, { "epoch": 4.16, "learning_rate": 4.63612722544509e-06, "loss": 0.012, "step": 13418 }, { "epoch": 4.16, "learning_rate": 4.635727145429087e-06, "loss": 0.0187, "step": 13419 }, { "epoch": 4.17, "learning_rate": 4.6353270654130824e-06, "loss": 0.0047, "step": 13420 }, { "epoch": 4.17, "learning_rate": 4.63492698539708e-06, "loss": 0.0193, "step": 13421 }, { "epoch": 4.17, "learning_rate": 4.634526905381077e-06, "loss": 0.0237, "step": 13422 }, { "epoch": 4.17, "learning_rate": 4.634126825365073e-06, "loss": 0.0303, "step": 13423 }, { "epoch": 4.17, "learning_rate": 4.63372674534907e-06, "loss": 0.0106, "step": 13424 }, { "epoch": 4.17, "learning_rate": 4.6333266653330676e-06, "loss": 0.0502, "step": 13425 }, { "epoch": 4.17, "learning_rate": 4.632926585317063e-06, "loss": 0.0151, "step": 13426 }, { "epoch": 4.17, "learning_rate": 4.63252650530106e-06, "loss": 0.0139, "step": 13427 }, { "epoch": 4.17, "learning_rate": 4.632126425285058e-06, "loss": 0.0089, "step": 13428 }, { "epoch": 4.17, "learning_rate": 4.631726345269054e-06, "loss": 0.0201, "step": 13429 }, { "epoch": 4.17, "learning_rate": 4.631326265253051e-06, "loss": 0.0111, "step": 13430 }, { "epoch": 4.17, "learning_rate": 4.630926185237048e-06, "loss": 0.0106, "step": 13431 }, { "epoch": 4.17, "learning_rate": 4.630526105221045e-06, "loss": 0.0054, "step": 13432 }, { "epoch": 4.17, "learning_rate": 4.630126025205041e-06, "loss": 0.0184, "step": 13433 }, { "epoch": 4.17, "learning_rate": 4.629725945189038e-06, "loss": 0.0476, "step": 13434 }, { "epoch": 4.17, "learning_rate": 4.629325865173035e-06, "loss": 0.0121, "step": 13435 }, { "epoch": 4.17, "learning_rate": 4.628925785157032e-06, "loss": 0.0217, "step": 13436 }, { "epoch": 4.17, "learning_rate": 4.628525705141029e-06, "loss": 0.007, "step": 13437 }, { "epoch": 4.17, "learning_rate": 4.628125625125025e-06, "loss": 0.0419, "step": 13438 }, { "epoch": 4.17, "learning_rate": 4.627725545109022e-06, "loss": 0.0085, "step": 13439 }, { "epoch": 4.17, "learning_rate": 4.627325465093019e-06, "loss": 0.0158, "step": 13440 }, { "epoch": 4.17, "learning_rate": 4.626925385077015e-06, "loss": 0.0111, "step": 13441 }, { "epoch": 4.17, "learning_rate": 4.626525305061013e-06, "loss": 0.0099, "step": 13442 }, { "epoch": 4.17, "learning_rate": 4.62612522504501e-06, "loss": 0.0749, "step": 13443 }, { "epoch": 4.17, "learning_rate": 4.625725145029006e-06, "loss": 0.0186, "step": 13444 }, { "epoch": 4.17, "learning_rate": 4.625325065013003e-06, "loss": 0.027, "step": 13445 }, { "epoch": 4.17, "learning_rate": 4.624924984997e-06, "loss": 0.0345, "step": 13446 }, { "epoch": 4.17, "learning_rate": 4.624524904980996e-06, "loss": 0.0368, "step": 13447 }, { "epoch": 4.17, "learning_rate": 4.624124824964993e-06, "loss": 0.0042, "step": 13448 }, { "epoch": 4.17, "learning_rate": 4.623724744948991e-06, "loss": 0.0387, "step": 13449 }, { "epoch": 4.17, "learning_rate": 4.623324664932987e-06, "loss": 0.0139, "step": 13450 }, { "epoch": 4.17, "learning_rate": 4.622924584916984e-06, "loss": 0.0178, "step": 13451 }, { "epoch": 4.18, "learning_rate": 4.622524504900981e-06, "loss": 0.016, "step": 13452 }, { "epoch": 4.18, "learning_rate": 4.622124424884977e-06, "loss": 0.0366, "step": 13453 }, { "epoch": 4.18, "learning_rate": 4.621724344868974e-06, "loss": 0.024, "step": 13454 }, { "epoch": 4.18, "learning_rate": 4.621324264852971e-06, "loss": 0.0199, "step": 13455 }, { "epoch": 4.18, "learning_rate": 4.620924184836968e-06, "loss": 0.113, "step": 13456 }, { "epoch": 4.18, "learning_rate": 4.620524104820965e-06, "loss": 0.0161, "step": 13457 }, { "epoch": 4.18, "learning_rate": 4.620124024804961e-06, "loss": 0.0137, "step": 13458 }, { "epoch": 4.18, "learning_rate": 4.619723944788958e-06, "loss": 0.0587, "step": 13459 }, { "epoch": 4.18, "learning_rate": 4.619323864772955e-06, "loss": 0.0118, "step": 13460 }, { "epoch": 4.18, "learning_rate": 4.618923784756952e-06, "loss": 0.0247, "step": 13461 }, { "epoch": 4.18, "learning_rate": 4.618523704740948e-06, "loss": 0.0143, "step": 13462 }, { "epoch": 4.18, "learning_rate": 4.618123624724946e-06, "loss": 0.0224, "step": 13463 }, { "epoch": 4.18, "learning_rate": 4.617723544708942e-06, "loss": 0.0457, "step": 13464 }, { "epoch": 4.18, "learning_rate": 4.617323464692938e-06, "loss": 0.067, "step": 13465 }, { "epoch": 4.18, "learning_rate": 4.616923384676936e-06, "loss": 0.058, "step": 13466 }, { "epoch": 4.18, "learning_rate": 4.616523304660933e-06, "loss": 0.0152, "step": 13467 }, { "epoch": 4.18, "learning_rate": 4.616123224644929e-06, "loss": 0.0201, "step": 13468 }, { "epoch": 4.18, "learning_rate": 4.615723144628926e-06, "loss": 0.0229, "step": 13469 }, { "epoch": 4.18, "learning_rate": 4.6153230646129235e-06, "loss": 0.0185, "step": 13470 }, { "epoch": 4.18, "learning_rate": 4.614922984596919e-06, "loss": 0.0089, "step": 13471 }, { "epoch": 4.18, "learning_rate": 4.614522904580916e-06, "loss": 0.0108, "step": 13472 }, { "epoch": 4.18, "learning_rate": 4.614122824564914e-06, "loss": 0.0058, "step": 13473 }, { "epoch": 4.18, "learning_rate": 4.61372274454891e-06, "loss": 0.0137, "step": 13474 }, { "epoch": 4.18, "learning_rate": 4.613322664532907e-06, "loss": 0.0165, "step": 13475 }, { "epoch": 4.18, "learning_rate": 4.612922584516904e-06, "loss": 0.0239, "step": 13476 }, { "epoch": 4.18, "learning_rate": 4.6125225045009e-06, "loss": 0.0173, "step": 13477 }, { "epoch": 4.18, "learning_rate": 4.612122424484897e-06, "loss": 0.0083, "step": 13478 }, { "epoch": 4.18, "learning_rate": 4.611722344468894e-06, "loss": 0.0238, "step": 13479 }, { "epoch": 4.18, "learning_rate": 4.611322264452891e-06, "loss": 0.0143, "step": 13480 }, { "epoch": 4.18, "learning_rate": 4.610922184436888e-06, "loss": 0.0502, "step": 13481 }, { "epoch": 4.18, "learning_rate": 4.610522104420885e-06, "loss": 0.0668, "step": 13482 }, { "epoch": 4.18, "learning_rate": 4.610122024404881e-06, "loss": 0.0072, "step": 13483 }, { "epoch": 4.19, "learning_rate": 4.609721944388878e-06, "loss": 0.0178, "step": 13484 }, { "epoch": 4.19, "learning_rate": 4.609321864372875e-06, "loss": 0.0061, "step": 13485 }, { "epoch": 4.19, "learning_rate": 4.608921784356871e-06, "loss": 0.0147, "step": 13486 }, { "epoch": 4.19, "learning_rate": 4.608521704340869e-06, "loss": 0.0067, "step": 13487 }, { "epoch": 4.19, "learning_rate": 4.608121624324866e-06, "loss": 0.0343, "step": 13488 }, { "epoch": 4.19, "learning_rate": 4.607721544308862e-06, "loss": 0.0084, "step": 13489 }, { "epoch": 4.19, "learning_rate": 4.607321464292859e-06, "loss": 0.0137, "step": 13490 }, { "epoch": 4.19, "learning_rate": 4.606921384276856e-06, "loss": 0.0699, "step": 13491 }, { "epoch": 4.19, "learning_rate": 4.606521304260852e-06, "loss": 0.0077, "step": 13492 }, { "epoch": 4.19, "learning_rate": 4.606121224244849e-06, "loss": 0.017, "step": 13493 }, { "epoch": 4.19, "learning_rate": 4.6057211442288465e-06, "loss": 0.0136, "step": 13494 }, { "epoch": 4.19, "learning_rate": 4.605321064212843e-06, "loss": 0.0312, "step": 13495 }, { "epoch": 4.19, "learning_rate": 4.60492098419684e-06, "loss": 0.0656, "step": 13496 }, { "epoch": 4.19, "learning_rate": 4.604520904180837e-06, "loss": 0.0153, "step": 13497 }, { "epoch": 4.19, "learning_rate": 4.604120824164833e-06, "loss": 0.0166, "step": 13498 }, { "epoch": 4.19, "learning_rate": 4.60372074414883e-06, "loss": 0.0186, "step": 13499 }, { "epoch": 4.19, "learning_rate": 4.603320664132827e-06, "loss": 0.0528, "step": 13500 }, { "epoch": 4.19, "learning_rate": 4.602920584116824e-06, "loss": 0.0331, "step": 13501 }, { "epoch": 4.19, "learning_rate": 4.60252050410082e-06, "loss": 0.0322, "step": 13502 }, { "epoch": 4.19, "learning_rate": 4.6021204240848176e-06, "loss": 0.1008, "step": 13503 }, { "epoch": 4.19, "learning_rate": 4.601720344068814e-06, "loss": 0.0078, "step": 13504 }, { "epoch": 4.19, "learning_rate": 4.601320264052811e-06, "loss": 0.0273, "step": 13505 }, { "epoch": 4.19, "learning_rate": 4.600920184036808e-06, "loss": 0.0365, "step": 13506 }, { "epoch": 4.19, "learning_rate": 4.600520104020804e-06, "loss": 0.0918, "step": 13507 }, { "epoch": 4.19, "learning_rate": 4.600120024004801e-06, "loss": 0.0133, "step": 13508 }, { "epoch": 4.19, "learning_rate": 4.599719943988798e-06, "loss": 0.0676, "step": 13509 }, { "epoch": 4.19, "learning_rate": 4.599319863972795e-06, "loss": 0.0215, "step": 13510 }, { "epoch": 4.19, "learning_rate": 4.598919783956792e-06, "loss": 0.0078, "step": 13511 }, { "epoch": 4.19, "learning_rate": 4.598519703940789e-06, "loss": 0.0452, "step": 13512 }, { "epoch": 4.19, "learning_rate": 4.598119623924785e-06, "loss": 0.0095, "step": 13513 }, { "epoch": 4.19, "learning_rate": 4.597719543908783e-06, "loss": 0.019, "step": 13514 }, { "epoch": 4.19, "learning_rate": 4.597319463892779e-06, "loss": 0.014, "step": 13515 }, { "epoch": 4.2, "learning_rate": 4.596919383876775e-06, "loss": 0.02, "step": 13516 }, { "epoch": 4.2, "learning_rate": 4.596519303860773e-06, "loss": 0.02, "step": 13517 }, { "epoch": 4.2, "learning_rate": 4.5961192238447696e-06, "loss": 0.0251, "step": 13518 }, { "epoch": 4.2, "learning_rate": 4.595719143828766e-06, "loss": 0.0145, "step": 13519 }, { "epoch": 4.2, "learning_rate": 4.595319063812763e-06, "loss": 0.0034, "step": 13520 }, { "epoch": 4.2, "learning_rate": 4.59491898379676e-06, "loss": 0.0399, "step": 13521 }, { "epoch": 4.2, "learning_rate": 4.594518903780756e-06, "loss": 0.0215, "step": 13522 }, { "epoch": 4.2, "learning_rate": 4.594118823764753e-06, "loss": 0.0132, "step": 13523 }, { "epoch": 4.2, "learning_rate": 4.5937187437487505e-06, "loss": 0.0137, "step": 13524 }, { "epoch": 4.2, "learning_rate": 4.593318663732747e-06, "loss": 0.038, "step": 13525 }, { "epoch": 4.2, "learning_rate": 4.592918583716744e-06, "loss": 0.0581, "step": 13526 }, { "epoch": 4.2, "learning_rate": 4.592518503700741e-06, "loss": 0.0092, "step": 13527 }, { "epoch": 4.2, "learning_rate": 4.592118423684737e-06, "loss": 0.0086, "step": 13528 }, { "epoch": 4.2, "learning_rate": 4.591718343668734e-06, "loss": 0.018, "step": 13529 }, { "epoch": 4.2, "learning_rate": 4.591318263652731e-06, "loss": 0.0357, "step": 13530 }, { "epoch": 4.2, "learning_rate": 4.590918183636728e-06, "loss": 0.017, "step": 13531 }, { "epoch": 4.2, "learning_rate": 4.590518103620725e-06, "loss": 0.03, "step": 13532 }, { "epoch": 4.2, "learning_rate": 4.5901180236047215e-06, "loss": 0.0077, "step": 13533 }, { "epoch": 4.2, "learning_rate": 4.589717943588718e-06, "loss": 0.0889, "step": 13534 }, { "epoch": 4.2, "learning_rate": 4.589317863572715e-06, "loss": 0.028, "step": 13535 }, { "epoch": 4.2, "learning_rate": 4.588917783556712e-06, "loss": 0.0205, "step": 13536 }, { "epoch": 4.2, "learning_rate": 4.588517703540708e-06, "loss": 0.0319, "step": 13537 }, { "epoch": 4.2, "learning_rate": 4.588117623524706e-06, "loss": 0.0184, "step": 13538 }, { "epoch": 4.2, "learning_rate": 4.5877175435087025e-06, "loss": 0.0092, "step": 13539 }, { "epoch": 4.2, "learning_rate": 4.587317463492698e-06, "loss": 0.0288, "step": 13540 }, { "epoch": 4.2, "learning_rate": 4.586917383476696e-06, "loss": 0.0202, "step": 13541 }, { "epoch": 4.2, "learning_rate": 4.5865173034606926e-06, "loss": 0.0254, "step": 13542 }, { "epoch": 4.2, "learning_rate": 4.586117223444689e-06, "loss": 0.0158, "step": 13543 }, { "epoch": 4.2, "learning_rate": 4.585717143428686e-06, "loss": 0.0228, "step": 13544 }, { "epoch": 4.2, "learning_rate": 4.5853170634126835e-06, "loss": 0.0135, "step": 13545 }, { "epoch": 4.2, "learning_rate": 4.584916983396679e-06, "loss": 0.028, "step": 13546 }, { "epoch": 4.2, "learning_rate": 4.584516903380676e-06, "loss": 0.0228, "step": 13547 }, { "epoch": 4.21, "learning_rate": 4.5841168233646735e-06, "loss": 0.0334, "step": 13548 }, { "epoch": 4.21, "learning_rate": 4.58371674334867e-06, "loss": 0.015, "step": 13549 }, { "epoch": 4.21, "learning_rate": 4.583316663332667e-06, "loss": 0.0239, "step": 13550 }, { "epoch": 4.21, "learning_rate": 4.582916583316664e-06, "loss": 0.1745, "step": 13551 }, { "epoch": 4.21, "learning_rate": 4.582516503300661e-06, "loss": 0.0218, "step": 13552 }, { "epoch": 4.21, "learning_rate": 4.582116423284657e-06, "loss": 0.0265, "step": 13553 }, { "epoch": 4.21, "learning_rate": 4.581716343268654e-06, "loss": 0.0099, "step": 13554 }, { "epoch": 4.21, "learning_rate": 4.581316263252651e-06, "loss": 0.0109, "step": 13555 }, { "epoch": 4.21, "learning_rate": 4.580916183236648e-06, "loss": 0.014, "step": 13556 }, { "epoch": 4.21, "learning_rate": 4.5805161032206446e-06, "loss": 0.0209, "step": 13557 }, { "epoch": 4.21, "learning_rate": 4.580116023204641e-06, "loss": 0.015, "step": 13558 }, { "epoch": 4.21, "learning_rate": 4.579715943188638e-06, "loss": 0.0175, "step": 13559 }, { "epoch": 4.21, "learning_rate": 4.579315863172635e-06, "loss": 0.0428, "step": 13560 }, { "epoch": 4.21, "learning_rate": 4.578915783156631e-06, "loss": 0.0548, "step": 13561 }, { "epoch": 4.21, "learning_rate": 4.578515703140629e-06, "loss": 0.0186, "step": 13562 }, { "epoch": 4.21, "learning_rate": 4.5781156231246255e-06, "loss": 0.0601, "step": 13563 }, { "epoch": 4.21, "learning_rate": 4.577715543108622e-06, "loss": 0.013, "step": 13564 }, { "epoch": 4.21, "learning_rate": 4.577315463092619e-06, "loss": 0.0157, "step": 13565 }, { "epoch": 4.21, "learning_rate": 4.576915383076616e-06, "loss": 0.0237, "step": 13566 }, { "epoch": 4.21, "learning_rate": 4.576515303060612e-06, "loss": 0.0075, "step": 13567 }, { "epoch": 4.21, "learning_rate": 4.576115223044609e-06, "loss": 0.0188, "step": 13568 }, { "epoch": 4.21, "learning_rate": 4.5757151430286065e-06, "loss": 0.0119, "step": 13569 }, { "epoch": 4.21, "learning_rate": 4.575315063012603e-06, "loss": 0.0498, "step": 13570 }, { "epoch": 4.21, "learning_rate": 4.574914982996599e-06, "loss": 0.0408, "step": 13571 }, { "epoch": 4.21, "learning_rate": 4.5745149029805965e-06, "loss": 0.0245, "step": 13572 }, { "epoch": 4.21, "learning_rate": 4.574114822964593e-06, "loss": 0.0438, "step": 13573 }, { "epoch": 4.21, "learning_rate": 4.57371474294859e-06, "loss": 0.0194, "step": 13574 }, { "epoch": 4.21, "learning_rate": 4.573314662932587e-06, "loss": 0.0381, "step": 13575 }, { "epoch": 4.21, "learning_rate": 4.572914582916584e-06, "loss": 0.0473, "step": 13576 }, { "epoch": 4.21, "learning_rate": 4.572514502900581e-06, "loss": 0.0206, "step": 13577 }, { "epoch": 4.21, "learning_rate": 4.572114422884577e-06, "loss": 0.018, "step": 13578 }, { "epoch": 4.21, "learning_rate": 4.571714342868574e-06, "loss": 0.0275, "step": 13579 }, { "epoch": 4.21, "learning_rate": 4.571314262852571e-06, "loss": 0.0099, "step": 13580 }, { "epoch": 4.22, "learning_rate": 4.5709141828365676e-06, "loss": 0.0145, "step": 13581 }, { "epoch": 4.22, "learning_rate": 4.570514102820564e-06, "loss": 0.015, "step": 13582 }, { "epoch": 4.22, "learning_rate": 4.570114022804562e-06, "loss": 0.0273, "step": 13583 }, { "epoch": 4.22, "learning_rate": 4.569713942788558e-06, "loss": 0.0315, "step": 13584 }, { "epoch": 4.22, "learning_rate": 4.569313862772554e-06, "loss": 0.0166, "step": 13585 }, { "epoch": 4.22, "learning_rate": 4.568913782756552e-06, "loss": 0.0163, "step": 13586 }, { "epoch": 4.22, "learning_rate": 4.5685137027405485e-06, "loss": 0.0178, "step": 13587 }, { "epoch": 4.22, "learning_rate": 4.568113622724545e-06, "loss": 0.0079, "step": 13588 }, { "epoch": 4.22, "learning_rate": 4.567713542708542e-06, "loss": 0.0312, "step": 13589 }, { "epoch": 4.22, "learning_rate": 4.567313462692539e-06, "loss": 0.0268, "step": 13590 }, { "epoch": 4.22, "learning_rate": 4.566913382676535e-06, "loss": 0.0213, "step": 13591 }, { "epoch": 4.22, "learning_rate": 4.566513302660533e-06, "loss": 0.014, "step": 13592 }, { "epoch": 4.22, "learning_rate": 4.5661132226445295e-06, "loss": 0.0081, "step": 13593 }, { "epoch": 4.22, "learning_rate": 4.565713142628526e-06, "loss": 0.0084, "step": 13594 }, { "epoch": 4.22, "learning_rate": 4.565313062612523e-06, "loss": 0.0115, "step": 13595 }, { "epoch": 4.22, "learning_rate": 4.5649129825965196e-06, "loss": 0.0072, "step": 13596 }, { "epoch": 4.22, "learning_rate": 4.564512902580516e-06, "loss": 0.0325, "step": 13597 }, { "epoch": 4.22, "learning_rate": 4.564112822564513e-06, "loss": 0.0118, "step": 13598 }, { "epoch": 4.22, "learning_rate": 4.5637127425485105e-06, "loss": 0.0224, "step": 13599 }, { "epoch": 4.22, "learning_rate": 4.563312662532507e-06, "loss": 0.0183, "step": 13600 }, { "epoch": 4.22, "learning_rate": 4.562912582516504e-06, "loss": 0.0196, "step": 13601 }, { "epoch": 4.22, "learning_rate": 4.5625125025005005e-06, "loss": 0.0145, "step": 13602 }, { "epoch": 4.22, "learning_rate": 4.562112422484497e-06, "loss": 0.0232, "step": 13603 }, { "epoch": 4.22, "learning_rate": 4.561712342468494e-06, "loss": 0.0378, "step": 13604 }, { "epoch": 4.22, "learning_rate": 4.561312262452491e-06, "loss": 0.0263, "step": 13605 }, { "epoch": 4.22, "learning_rate": 4.560912182436488e-06, "loss": 0.0161, "step": 13606 }, { "epoch": 4.22, "learning_rate": 4.560512102420485e-06, "loss": 0.0072, "step": 13607 }, { "epoch": 4.22, "learning_rate": 4.5601120224044815e-06, "loss": 0.0075, "step": 13608 }, { "epoch": 4.22, "learning_rate": 4.559711942388478e-06, "loss": 0.0306, "step": 13609 }, { "epoch": 4.22, "learning_rate": 4.559311862372475e-06, "loss": 0.0338, "step": 13610 }, { "epoch": 4.22, "learning_rate": 4.5589117823564715e-06, "loss": 0.0097, "step": 13611 }, { "epoch": 4.22, "learning_rate": 4.558511702340468e-06, "loss": 0.0108, "step": 13612 }, { "epoch": 4.23, "learning_rate": 4.558111622324466e-06, "loss": 0.011, "step": 13613 }, { "epoch": 4.23, "learning_rate": 4.5577115423084624e-06, "loss": 0.0136, "step": 13614 }, { "epoch": 4.23, "learning_rate": 4.557311462292459e-06, "loss": 0.011, "step": 13615 }, { "epoch": 4.23, "learning_rate": 4.556911382276456e-06, "loss": 0.015, "step": 13616 }, { "epoch": 4.23, "learning_rate": 4.5565113022604525e-06, "loss": 0.0115, "step": 13617 }, { "epoch": 4.23, "learning_rate": 4.556111222244449e-06, "loss": 0.0209, "step": 13618 }, { "epoch": 4.23, "learning_rate": 4.555711142228446e-06, "loss": 0.0472, "step": 13619 }, { "epoch": 4.23, "learning_rate": 4.555311062212443e-06, "loss": 0.0057, "step": 13620 }, { "epoch": 4.23, "learning_rate": 4.55491098219644e-06, "loss": 0.0897, "step": 13621 }, { "epoch": 4.23, "learning_rate": 4.554510902180436e-06, "loss": 0.0746, "step": 13622 }, { "epoch": 4.23, "learning_rate": 4.5541108221644335e-06, "loss": 0.0126, "step": 13623 }, { "epoch": 4.23, "learning_rate": 4.55371074214843e-06, "loss": 0.0101, "step": 13624 }, { "epoch": 4.23, "learning_rate": 4.553310662132427e-06, "loss": 0.1484, "step": 13625 }, { "epoch": 4.23, "learning_rate": 4.5529105821164235e-06, "loss": 0.0096, "step": 13626 }, { "epoch": 4.23, "learning_rate": 4.552510502100421e-06, "loss": 0.0143, "step": 13627 }, { "epoch": 4.23, "learning_rate": 4.552110422084417e-06, "loss": 0.0669, "step": 13628 }, { "epoch": 4.23, "learning_rate": 4.551710342068414e-06, "loss": 0.0142, "step": 13629 }, { "epoch": 4.23, "learning_rate": 4.551310262052411e-06, "loss": 0.02, "step": 13630 }, { "epoch": 4.23, "learning_rate": 4.550910182036408e-06, "loss": 0.0446, "step": 13631 }, { "epoch": 4.23, "learning_rate": 4.5505101020204045e-06, "loss": 0.01, "step": 13632 }, { "epoch": 4.23, "learning_rate": 4.550110022004401e-06, "loss": 0.0123, "step": 13633 }, { "epoch": 4.23, "learning_rate": 4.549709941988399e-06, "loss": 0.0294, "step": 13634 }, { "epoch": 4.23, "learning_rate": 4.5493098619723946e-06, "loss": 0.0302, "step": 13635 }, { "epoch": 4.23, "learning_rate": 4.548909781956391e-06, "loss": 0.0066, "step": 13636 }, { "epoch": 4.23, "learning_rate": 4.548509701940389e-06, "loss": 0.0058, "step": 13637 }, { "epoch": 4.23, "learning_rate": 4.5481096219243855e-06, "loss": 0.0519, "step": 13638 }, { "epoch": 4.23, "learning_rate": 4.547709541908382e-06, "loss": 0.0146, "step": 13639 }, { "epoch": 4.23, "learning_rate": 4.547309461892379e-06, "loss": 0.016, "step": 13640 }, { "epoch": 4.23, "learning_rate": 4.5469093818763755e-06, "loss": 0.027, "step": 13641 }, { "epoch": 4.23, "learning_rate": 4.546509301860372e-06, "loss": 0.0272, "step": 13642 }, { "epoch": 4.23, "learning_rate": 4.546109221844369e-06, "loss": 0.0352, "step": 13643 }, { "epoch": 4.23, "learning_rate": 4.5457091418283664e-06, "loss": 0.1428, "step": 13644 }, { "epoch": 4.24, "learning_rate": 4.545309061812363e-06, "loss": 0.0423, "step": 13645 }, { "epoch": 4.24, "learning_rate": 4.54490898179636e-06, "loss": 0.0096, "step": 13646 }, { "epoch": 4.24, "learning_rate": 4.5445089017803565e-06, "loss": 0.0328, "step": 13647 }, { "epoch": 4.24, "learning_rate": 4.544108821764353e-06, "loss": 0.0117, "step": 13648 }, { "epoch": 4.24, "learning_rate": 4.54370874174835e-06, "loss": 0.0279, "step": 13649 }, { "epoch": 4.24, "learning_rate": 4.5433086617323465e-06, "loss": 0.006, "step": 13650 }, { "epoch": 4.24, "learning_rate": 4.542908581716344e-06, "loss": 0.0126, "step": 13651 }, { "epoch": 4.24, "learning_rate": 4.542508501700341e-06, "loss": 0.0093, "step": 13652 }, { "epoch": 4.24, "learning_rate": 4.542108421684337e-06, "loss": 0.0183, "step": 13653 }, { "epoch": 4.24, "learning_rate": 4.541708341668334e-06, "loss": 0.0197, "step": 13654 }, { "epoch": 4.24, "learning_rate": 4.541308261652331e-06, "loss": 0.0248, "step": 13655 }, { "epoch": 4.24, "learning_rate": 4.5409081816363275e-06, "loss": 0.0345, "step": 13656 }, { "epoch": 4.24, "learning_rate": 4.540508101620324e-06, "loss": 0.0064, "step": 13657 }, { "epoch": 4.24, "learning_rate": 4.540108021604322e-06, "loss": 0.021, "step": 13658 }, { "epoch": 4.24, "learning_rate": 4.539707941588318e-06, "loss": 0.0117, "step": 13659 }, { "epoch": 4.24, "learning_rate": 4.539307861572314e-06, "loss": 0.0129, "step": 13660 }, { "epoch": 4.24, "learning_rate": 4.538907781556312e-06, "loss": 0.0126, "step": 13661 }, { "epoch": 4.24, "learning_rate": 4.5385077015403085e-06, "loss": 0.039, "step": 13662 }, { "epoch": 4.24, "learning_rate": 4.538107621524305e-06, "loss": 0.0256, "step": 13663 }, { "epoch": 4.24, "learning_rate": 4.537707541508302e-06, "loss": 0.0653, "step": 13664 }, { "epoch": 4.24, "learning_rate": 4.537307461492299e-06, "loss": 0.0097, "step": 13665 }, { "epoch": 4.24, "learning_rate": 4.536907381476295e-06, "loss": 0.0289, "step": 13666 }, { "epoch": 4.24, "learning_rate": 4.536507301460292e-06, "loss": 0.0226, "step": 13667 }, { "epoch": 4.24, "learning_rate": 4.5361072214442894e-06, "loss": 0.0071, "step": 13668 }, { "epoch": 4.24, "learning_rate": 4.535707141428286e-06, "loss": 0.0359, "step": 13669 }, { "epoch": 4.24, "learning_rate": 4.535307061412283e-06, "loss": 0.0114, "step": 13670 }, { "epoch": 4.24, "learning_rate": 4.5349069813962795e-06, "loss": 0.0127, "step": 13671 }, { "epoch": 4.24, "learning_rate": 4.534506901380276e-06, "loss": 0.007, "step": 13672 }, { "epoch": 4.24, "learning_rate": 4.534106821364273e-06, "loss": 0.0108, "step": 13673 }, { "epoch": 4.24, "learning_rate": 4.5337067413482696e-06, "loss": 0.0214, "step": 13674 }, { "epoch": 4.24, "learning_rate": 4.533306661332267e-06, "loss": 0.0143, "step": 13675 }, { "epoch": 4.24, "learning_rate": 4.532906581316264e-06, "loss": 0.0252, "step": 13676 }, { "epoch": 4.25, "learning_rate": 4.5325065013002605e-06, "loss": 0.0659, "step": 13677 }, { "epoch": 4.25, "learning_rate": 4.532106421284257e-06, "loss": 0.0078, "step": 13678 }, { "epoch": 4.25, "learning_rate": 4.531706341268254e-06, "loss": 0.0388, "step": 13679 }, { "epoch": 4.25, "learning_rate": 4.5313062612522505e-06, "loss": 0.0108, "step": 13680 }, { "epoch": 4.25, "learning_rate": 4.530906181236247e-06, "loss": 0.0318, "step": 13681 }, { "epoch": 4.25, "learning_rate": 4.530506101220245e-06, "loss": 0.0232, "step": 13682 }, { "epoch": 4.25, "learning_rate": 4.5301060212042414e-06, "loss": 0.0127, "step": 13683 }, { "epoch": 4.25, "learning_rate": 4.529705941188238e-06, "loss": 0.0362, "step": 13684 }, { "epoch": 4.25, "learning_rate": 4.529305861172235e-06, "loss": 0.0215, "step": 13685 }, { "epoch": 4.25, "learning_rate": 4.5289057811562315e-06, "loss": 0.03, "step": 13686 }, { "epoch": 4.25, "learning_rate": 4.528505701140228e-06, "loss": 0.0162, "step": 13687 }, { "epoch": 4.25, "learning_rate": 4.528105621124225e-06, "loss": 0.0264, "step": 13688 }, { "epoch": 4.25, "learning_rate": 4.527705541108222e-06, "loss": 0.0176, "step": 13689 }, { "epoch": 4.25, "learning_rate": 4.527305461092219e-06, "loss": 0.111, "step": 13690 }, { "epoch": 4.25, "learning_rate": 4.526905381076216e-06, "loss": 0.012, "step": 13691 }, { "epoch": 4.25, "learning_rate": 4.5265053010602124e-06, "loss": 0.0131, "step": 13692 }, { "epoch": 4.25, "learning_rate": 4.526105221044209e-06, "loss": 0.0262, "step": 13693 }, { "epoch": 4.25, "learning_rate": 4.525705141028206e-06, "loss": 0.0189, "step": 13694 }, { "epoch": 4.25, "learning_rate": 4.5253050610122025e-06, "loss": 0.0501, "step": 13695 }, { "epoch": 4.25, "learning_rate": 4.5249049809962e-06, "loss": 0.0127, "step": 13696 }, { "epoch": 4.25, "learning_rate": 4.524504900980197e-06, "loss": 0.015, "step": 13697 }, { "epoch": 4.25, "learning_rate": 4.524104820964193e-06, "loss": 0.0051, "step": 13698 }, { "epoch": 4.25, "learning_rate": 4.52370474094819e-06, "loss": 0.0184, "step": 13699 }, { "epoch": 4.25, "learning_rate": 4.523304660932187e-06, "loss": 0.0349, "step": 13700 }, { "epoch": 4.25, "learning_rate": 4.5229045809161835e-06, "loss": 0.0587, "step": 13701 }, { "epoch": 4.25, "learning_rate": 4.52250450090018e-06, "loss": 0.1359, "step": 13702 }, { "epoch": 4.25, "learning_rate": 4.522104420884178e-06, "loss": 0.0325, "step": 13703 }, { "epoch": 4.25, "learning_rate": 4.5217043408681735e-06, "loss": 0.0378, "step": 13704 }, { "epoch": 4.25, "learning_rate": 4.521304260852171e-06, "loss": 0.0141, "step": 13705 }, { "epoch": 4.25, "learning_rate": 4.520904180836168e-06, "loss": 0.018, "step": 13706 }, { "epoch": 4.25, "learning_rate": 4.5205041008201644e-06, "loss": 0.0071, "step": 13707 }, { "epoch": 4.25, "learning_rate": 4.520104020804161e-06, "loss": 0.0436, "step": 13708 }, { "epoch": 4.25, "learning_rate": 4.519703940788158e-06, "loss": 0.0076, "step": 13709 }, { "epoch": 4.26, "learning_rate": 4.5193038607721545e-06, "loss": 0.0087, "step": 13710 }, { "epoch": 4.26, "learning_rate": 4.518903780756151e-06, "loss": 0.0353, "step": 13711 }, { "epoch": 4.26, "learning_rate": 4.518503700740149e-06, "loss": 0.0434, "step": 13712 }, { "epoch": 4.26, "learning_rate": 4.518103620724145e-06, "loss": 0.0101, "step": 13713 }, { "epoch": 4.26, "learning_rate": 4.517703540708142e-06, "loss": 0.0099, "step": 13714 }, { "epoch": 4.26, "learning_rate": 4.517303460692139e-06, "loss": 0.0209, "step": 13715 }, { "epoch": 4.26, "learning_rate": 4.5169033806761355e-06, "loss": 0.0243, "step": 13716 }, { "epoch": 4.26, "learning_rate": 4.516503300660132e-06, "loss": 0.0106, "step": 13717 }, { "epoch": 4.26, "learning_rate": 4.516103220644129e-06, "loss": 0.0112, "step": 13718 }, { "epoch": 4.26, "learning_rate": 4.515703140628126e-06, "loss": 0.0195, "step": 13719 }, { "epoch": 4.26, "learning_rate": 4.515303060612123e-06, "loss": 0.0325, "step": 13720 }, { "epoch": 4.26, "learning_rate": 4.51490298059612e-06, "loss": 0.0267, "step": 13721 }, { "epoch": 4.26, "learning_rate": 4.5145029005801164e-06, "loss": 0.0528, "step": 13722 }, { "epoch": 4.26, "learning_rate": 4.514102820564113e-06, "loss": 0.0217, "step": 13723 }, { "epoch": 4.26, "learning_rate": 4.51370274054811e-06, "loss": 0.0294, "step": 13724 }, { "epoch": 4.26, "learning_rate": 4.5133026605321065e-06, "loss": 0.0189, "step": 13725 }, { "epoch": 4.26, "learning_rate": 4.512902580516104e-06, "loss": 0.0255, "step": 13726 }, { "epoch": 4.26, "learning_rate": 4.512502500500101e-06, "loss": 0.0255, "step": 13727 }, { "epoch": 4.26, "learning_rate": 4.512102420484097e-06, "loss": 0.0245, "step": 13728 }, { "epoch": 4.26, "learning_rate": 4.511702340468094e-06, "loss": 0.0265, "step": 13729 }, { "epoch": 4.26, "learning_rate": 4.511302260452091e-06, "loss": 0.0103, "step": 13730 }, { "epoch": 4.26, "learning_rate": 4.5109021804360874e-06, "loss": 0.0219, "step": 13731 }, { "epoch": 4.26, "learning_rate": 4.510502100420084e-06, "loss": 0.0085, "step": 13732 }, { "epoch": 4.26, "learning_rate": 4.510102020404082e-06, "loss": 0.0201, "step": 13733 }, { "epoch": 4.26, "learning_rate": 4.509701940388078e-06, "loss": 0.0081, "step": 13734 }, { "epoch": 4.26, "learning_rate": 4.509301860372074e-06, "loss": 0.0209, "step": 13735 }, { "epoch": 4.26, "learning_rate": 4.508901780356072e-06, "loss": 0.0278, "step": 13736 }, { "epoch": 4.26, "learning_rate": 4.508501700340068e-06, "loss": 0.015, "step": 13737 }, { "epoch": 4.26, "learning_rate": 4.508101620324065e-06, "loss": 0.0816, "step": 13738 }, { "epoch": 4.26, "learning_rate": 4.507701540308062e-06, "loss": 0.031, "step": 13739 }, { "epoch": 4.26, "learning_rate": 4.507301460292059e-06, "loss": 0.0117, "step": 13740 }, { "epoch": 4.26, "learning_rate": 4.506901380276056e-06, "loss": 0.0845, "step": 13741 }, { "epoch": 4.27, "learning_rate": 4.506501300260052e-06, "loss": 0.0112, "step": 13742 }, { "epoch": 4.27, "learning_rate": 4.506101220244049e-06, "loss": 0.0193, "step": 13743 }, { "epoch": 4.27, "learning_rate": 4.505701140228046e-06, "loss": 0.0048, "step": 13744 }, { "epoch": 4.27, "learning_rate": 4.505301060212043e-06, "loss": 0.0077, "step": 13745 }, { "epoch": 4.27, "learning_rate": 4.5049009801960394e-06, "loss": 0.0034, "step": 13746 }, { "epoch": 4.27, "learning_rate": 4.504500900180037e-06, "loss": 0.0136, "step": 13747 }, { "epoch": 4.27, "learning_rate": 4.504100820164033e-06, "loss": 0.011, "step": 13748 }, { "epoch": 4.27, "learning_rate": 4.5037007401480295e-06, "loss": 0.0184, "step": 13749 }, { "epoch": 4.27, "learning_rate": 4.503300660132027e-06, "loss": 0.0083, "step": 13750 }, { "epoch": 4.27, "learning_rate": 4.502900580116024e-06, "loss": 0.0199, "step": 13751 }, { "epoch": 4.27, "learning_rate": 4.50250050010002e-06, "loss": 0.0046, "step": 13752 }, { "epoch": 4.27, "learning_rate": 4.502100420084017e-06, "loss": 0.0099, "step": 13753 }, { "epoch": 4.27, "learning_rate": 4.501700340068014e-06, "loss": 0.0121, "step": 13754 }, { "epoch": 4.27, "learning_rate": 4.5013002600520105e-06, "loss": 0.0121, "step": 13755 }, { "epoch": 4.27, "learning_rate": 4.500900180036007e-06, "loss": 0.0323, "step": 13756 }, { "epoch": 4.27, "learning_rate": 4.500500100020005e-06, "loss": 0.0124, "step": 13757 }, { "epoch": 4.27, "learning_rate": 4.500100020004001e-06, "loss": 0.0068, "step": 13758 }, { "epoch": 4.27, "learning_rate": 4.499699939987998e-06, "loss": 0.0801, "step": 13759 }, { "epoch": 4.27, "learning_rate": 4.499299859971995e-06, "loss": 0.0564, "step": 13760 }, { "epoch": 4.27, "learning_rate": 4.4988997799559914e-06, "loss": 0.0138, "step": 13761 }, { "epoch": 4.27, "learning_rate": 4.498499699939988e-06, "loss": 0.0227, "step": 13762 }, { "epoch": 4.27, "learning_rate": 4.498099619923985e-06, "loss": 0.0267, "step": 13763 }, { "epoch": 4.27, "learning_rate": 4.497699539907982e-06, "loss": 0.0314, "step": 13764 }, { "epoch": 4.27, "learning_rate": 4.497299459891979e-06, "loss": 0.0257, "step": 13765 }, { "epoch": 4.27, "learning_rate": 4.496899379875976e-06, "loss": 0.0212, "step": 13766 }, { "epoch": 4.27, "learning_rate": 4.496499299859972e-06, "loss": 0.011, "step": 13767 }, { "epoch": 4.27, "learning_rate": 4.496099219843969e-06, "loss": 0.0444, "step": 13768 }, { "epoch": 4.27, "learning_rate": 4.495699139827966e-06, "loss": 0.0214, "step": 13769 }, { "epoch": 4.27, "learning_rate": 4.4952990598119624e-06, "loss": 0.0207, "step": 13770 }, { "epoch": 4.27, "learning_rate": 4.49489897979596e-06, "loss": 0.0252, "step": 13771 }, { "epoch": 4.27, "learning_rate": 4.494498899779957e-06, "loss": 0.0351, "step": 13772 }, { "epoch": 4.27, "learning_rate": 4.4940988197639525e-06, "loss": 0.0127, "step": 13773 }, { "epoch": 4.28, "learning_rate": 4.49369873974795e-06, "loss": 0.0066, "step": 13774 }, { "epoch": 4.28, "learning_rate": 4.493298659731947e-06, "loss": 0.0374, "step": 13775 }, { "epoch": 4.28, "learning_rate": 4.492898579715943e-06, "loss": 0.0107, "step": 13776 }, { "epoch": 4.28, "learning_rate": 4.49249849969994e-06, "loss": 0.0199, "step": 13777 }, { "epoch": 4.28, "learning_rate": 4.492098419683938e-06, "loss": 0.0221, "step": 13778 }, { "epoch": 4.28, "learning_rate": 4.491698339667934e-06, "loss": 0.1243, "step": 13779 }, { "epoch": 4.28, "learning_rate": 4.49129825965193e-06, "loss": 0.0218, "step": 13780 }, { "epoch": 4.28, "learning_rate": 4.490898179635928e-06, "loss": 0.0081, "step": 13781 }, { "epoch": 4.28, "learning_rate": 4.490498099619924e-06, "loss": 0.0577, "step": 13782 }, { "epoch": 4.28, "learning_rate": 4.490098019603921e-06, "loss": 0.0203, "step": 13783 }, { "epoch": 4.28, "learning_rate": 4.489697939587918e-06, "loss": 0.0374, "step": 13784 }, { "epoch": 4.28, "learning_rate": 4.489297859571915e-06, "loss": 0.0209, "step": 13785 }, { "epoch": 4.28, "learning_rate": 4.488897779555911e-06, "loss": 0.0182, "step": 13786 }, { "epoch": 4.28, "learning_rate": 4.488497699539908e-06, "loss": 0.0279, "step": 13787 }, { "epoch": 4.28, "learning_rate": 4.488097619523905e-06, "loss": 0.0156, "step": 13788 }, { "epoch": 4.28, "learning_rate": 4.487697539507902e-06, "loss": 0.1027, "step": 13789 }, { "epoch": 4.28, "learning_rate": 4.487297459491899e-06, "loss": 0.006, "step": 13790 }, { "epoch": 4.28, "learning_rate": 4.486897379475895e-06, "loss": 0.0196, "step": 13791 }, { "epoch": 4.28, "learning_rate": 4.486497299459892e-06, "loss": 0.0179, "step": 13792 }, { "epoch": 4.28, "learning_rate": 4.486097219443889e-06, "loss": 0.0936, "step": 13793 }, { "epoch": 4.28, "learning_rate": 4.4856971394278855e-06, "loss": 0.0198, "step": 13794 }, { "epoch": 4.28, "learning_rate": 4.485297059411883e-06, "loss": 0.009, "step": 13795 }, { "epoch": 4.28, "learning_rate": 4.48489697939588e-06, "loss": 0.0479, "step": 13796 }, { "epoch": 4.28, "learning_rate": 4.484496899379876e-06, "loss": 0.0313, "step": 13797 }, { "epoch": 4.28, "learning_rate": 4.484096819363873e-06, "loss": 0.0089, "step": 13798 }, { "epoch": 4.28, "learning_rate": 4.48369673934787e-06, "loss": 0.0055, "step": 13799 }, { "epoch": 4.28, "learning_rate": 4.4832966593318664e-06, "loss": 0.0204, "step": 13800 }, { "epoch": 4.28, "learning_rate": 4.482896579315863e-06, "loss": 0.0617, "step": 13801 }, { "epoch": 4.28, "learning_rate": 4.482496499299861e-06, "loss": 0.0121, "step": 13802 }, { "epoch": 4.28, "learning_rate": 4.482096419283857e-06, "loss": 0.0277, "step": 13803 }, { "epoch": 4.28, "learning_rate": 4.481696339267854e-06, "loss": 0.006, "step": 13804 }, { "epoch": 4.28, "learning_rate": 4.481296259251851e-06, "loss": 0.0199, "step": 13805 }, { "epoch": 4.29, "learning_rate": 4.480896179235847e-06, "loss": 0.0125, "step": 13806 }, { "epoch": 4.29, "learning_rate": 4.480496099219844e-06, "loss": 0.0266, "step": 13807 }, { "epoch": 4.29, "learning_rate": 4.480096019203841e-06, "loss": 0.0548, "step": 13808 }, { "epoch": 4.29, "learning_rate": 4.479695939187838e-06, "loss": 0.0165, "step": 13809 }, { "epoch": 4.29, "learning_rate": 4.479295859171835e-06, "loss": 0.0071, "step": 13810 }, { "epoch": 4.29, "learning_rate": 4.478895779155832e-06, "loss": 0.0175, "step": 13811 }, { "epoch": 4.29, "learning_rate": 4.478495699139828e-06, "loss": 0.1338, "step": 13812 }, { "epoch": 4.29, "learning_rate": 4.478095619123825e-06, "loss": 0.0305, "step": 13813 }, { "epoch": 4.29, "learning_rate": 4.477695539107822e-06, "loss": 0.0174, "step": 13814 }, { "epoch": 4.29, "learning_rate": 4.477295459091819e-06, "loss": 0.0471, "step": 13815 }, { "epoch": 4.29, "learning_rate": 4.476895379075816e-06, "loss": 0.0413, "step": 13816 }, { "epoch": 4.29, "learning_rate": 4.476495299059812e-06, "loss": 0.0151, "step": 13817 }, { "epoch": 4.29, "learning_rate": 4.476095219043809e-06, "loss": 0.0094, "step": 13818 }, { "epoch": 4.29, "learning_rate": 4.475695139027806e-06, "loss": 0.0213, "step": 13819 }, { "epoch": 4.29, "learning_rate": 4.475295059011803e-06, "loss": 0.01, "step": 13820 }, { "epoch": 4.29, "learning_rate": 4.474894978995799e-06, "loss": 0.0171, "step": 13821 }, { "epoch": 4.29, "learning_rate": 4.474494898979797e-06, "loss": 0.0273, "step": 13822 }, { "epoch": 4.29, "learning_rate": 4.474094818963794e-06, "loss": 0.0185, "step": 13823 }, { "epoch": 4.29, "learning_rate": 4.4736947389477894e-06, "loss": 0.0301, "step": 13824 }, { "epoch": 4.29, "learning_rate": 4.473294658931787e-06, "loss": 0.0551, "step": 13825 }, { "epoch": 4.29, "learning_rate": 4.472894578915784e-06, "loss": 0.0534, "step": 13826 }, { "epoch": 4.29, "learning_rate": 4.47249449889978e-06, "loss": 0.0118, "step": 13827 }, { "epoch": 4.29, "learning_rate": 4.472094418883777e-06, "loss": 0.0329, "step": 13828 }, { "epoch": 4.29, "learning_rate": 4.4716943388677746e-06, "loss": 0.0083, "step": 13829 }, { "epoch": 4.29, "learning_rate": 4.47129425885177e-06, "loss": 0.0139, "step": 13830 }, { "epoch": 4.29, "learning_rate": 4.470894178835767e-06, "loss": 0.007, "step": 13831 }, { "epoch": 4.29, "learning_rate": 4.470494098819765e-06, "loss": 0.0206, "step": 13832 }, { "epoch": 4.29, "learning_rate": 4.470094018803761e-06, "loss": 0.0243, "step": 13833 }, { "epoch": 4.29, "learning_rate": 4.469693938787758e-06, "loss": 0.0248, "step": 13834 }, { "epoch": 4.29, "learning_rate": 4.469293858771755e-06, "loss": 0.0154, "step": 13835 }, { "epoch": 4.29, "learning_rate": 4.468893778755751e-06, "loss": 0.042, "step": 13836 }, { "epoch": 4.29, "learning_rate": 4.468493698739748e-06, "loss": 0.0124, "step": 13837 }, { "epoch": 4.3, "learning_rate": 4.468093618723745e-06, "loss": 0.0119, "step": 13838 }, { "epoch": 4.3, "learning_rate": 4.467693538707742e-06, "loss": 0.021, "step": 13839 }, { "epoch": 4.3, "learning_rate": 4.467293458691739e-06, "loss": 0.0027, "step": 13840 }, { "epoch": 4.3, "learning_rate": 4.466893378675736e-06, "loss": 0.1018, "step": 13841 }, { "epoch": 4.3, "learning_rate": 4.466493298659732e-06, "loss": 0.0038, "step": 13842 }, { "epoch": 4.3, "learning_rate": 4.466093218643729e-06, "loss": 0.0184, "step": 13843 }, { "epoch": 4.3, "learning_rate": 4.465693138627726e-06, "loss": 0.0094, "step": 13844 }, { "epoch": 4.3, "learning_rate": 4.465293058611722e-06, "loss": 0.0164, "step": 13845 }, { "epoch": 4.3, "learning_rate": 4.46489297859572e-06, "loss": 0.0072, "step": 13846 }, { "epoch": 4.3, "learning_rate": 4.464492898579717e-06, "loss": 0.0201, "step": 13847 }, { "epoch": 4.3, "learning_rate": 4.464092818563713e-06, "loss": 0.0386, "step": 13848 }, { "epoch": 4.3, "learning_rate": 4.46369273854771e-06, "loss": 0.0363, "step": 13849 }, { "epoch": 4.3, "learning_rate": 4.463292658531707e-06, "loss": 0.0142, "step": 13850 }, { "epoch": 4.3, "learning_rate": 4.462892578515703e-06, "loss": 0.0173, "step": 13851 }, { "epoch": 4.3, "learning_rate": 4.4624924984997e-06, "loss": 0.0163, "step": 13852 }, { "epoch": 4.3, "learning_rate": 4.4620924184836976e-06, "loss": 0.0106, "step": 13853 }, { "epoch": 4.3, "learning_rate": 4.461692338467694e-06, "loss": 0.0326, "step": 13854 }, { "epoch": 4.3, "learning_rate": 4.46129225845169e-06, "loss": 0.0111, "step": 13855 }, { "epoch": 4.3, "learning_rate": 4.460892178435688e-06, "loss": 0.0204, "step": 13856 }, { "epoch": 4.3, "learning_rate": 4.460492098419684e-06, "loss": 0.022, "step": 13857 }, { "epoch": 4.3, "learning_rate": 4.460092018403681e-06, "loss": 0.0223, "step": 13858 }, { "epoch": 4.3, "learning_rate": 4.459691938387678e-06, "loss": 0.0124, "step": 13859 }, { "epoch": 4.3, "learning_rate": 4.459291858371675e-06, "loss": 0.0133, "step": 13860 }, { "epoch": 4.3, "learning_rate": 4.458891778355672e-06, "loss": 0.0164, "step": 13861 }, { "epoch": 4.3, "learning_rate": 4.458491698339668e-06, "loss": 0.0271, "step": 13862 }, { "epoch": 4.3, "learning_rate": 4.458091618323665e-06, "loss": 0.0075, "step": 13863 }, { "epoch": 4.3, "learning_rate": 4.457691538307662e-06, "loss": 0.0386, "step": 13864 }, { "epoch": 4.3, "learning_rate": 4.457291458291659e-06, "loss": 0.0128, "step": 13865 }, { "epoch": 4.3, "learning_rate": 4.456891378275655e-06, "loss": 0.0197, "step": 13866 }, { "epoch": 4.3, "learning_rate": 4.456491298259653e-06, "loss": 0.0272, "step": 13867 }, { "epoch": 4.3, "learning_rate": 4.456091218243649e-06, "loss": 0.0051, "step": 13868 }, { "epoch": 4.3, "learning_rate": 4.455691138227645e-06, "loss": 0.0293, "step": 13869 }, { "epoch": 4.3, "learning_rate": 4.455291058211643e-06, "loss": 0.0245, "step": 13870 }, { "epoch": 4.31, "learning_rate": 4.45489097819564e-06, "loss": 0.0126, "step": 13871 }, { "epoch": 4.31, "learning_rate": 4.454490898179636e-06, "loss": 0.0131, "step": 13872 }, { "epoch": 4.31, "learning_rate": 4.454090818163633e-06, "loss": 0.0113, "step": 13873 }, { "epoch": 4.31, "learning_rate": 4.45369073814763e-06, "loss": 0.0102, "step": 13874 }, { "epoch": 4.31, "learning_rate": 4.453290658131626e-06, "loss": 0.0092, "step": 13875 }, { "epoch": 4.31, "learning_rate": 4.452890578115623e-06, "loss": 0.0128, "step": 13876 }, { "epoch": 4.31, "learning_rate": 4.452490498099621e-06, "loss": 0.0159, "step": 13877 }, { "epoch": 4.31, "learning_rate": 4.452090418083617e-06, "loss": 0.0186, "step": 13878 }, { "epoch": 4.31, "learning_rate": 4.451690338067614e-06, "loss": 0.0103, "step": 13879 }, { "epoch": 4.31, "learning_rate": 4.451290258051611e-06, "loss": 0.0217, "step": 13880 }, { "epoch": 4.31, "learning_rate": 4.450890178035607e-06, "loss": 0.0208, "step": 13881 }, { "epoch": 4.31, "learning_rate": 4.450490098019604e-06, "loss": 0.0483, "step": 13882 }, { "epoch": 4.31, "learning_rate": 4.450090018003601e-06, "loss": 0.0177, "step": 13883 }, { "epoch": 4.31, "learning_rate": 4.449689937987598e-06, "loss": 0.013, "step": 13884 }, { "epoch": 4.31, "learning_rate": 4.449289857971595e-06, "loss": 0.0168, "step": 13885 }, { "epoch": 4.31, "learning_rate": 4.448889777955592e-06, "loss": 0.0069, "step": 13886 }, { "epoch": 4.31, "learning_rate": 4.448489697939588e-06, "loss": 0.0072, "step": 13887 }, { "epoch": 4.31, "learning_rate": 4.448089617923585e-06, "loss": 0.0139, "step": 13888 }, { "epoch": 4.31, "learning_rate": 4.447689537907582e-06, "loss": 0.0086, "step": 13889 }, { "epoch": 4.31, "learning_rate": 4.447289457891578e-06, "loss": 0.0149, "step": 13890 }, { "epoch": 4.31, "learning_rate": 4.446889377875576e-06, "loss": 0.0398, "step": 13891 }, { "epoch": 4.31, "learning_rate": 4.4464892978595726e-06, "loss": 0.0766, "step": 13892 }, { "epoch": 4.31, "learning_rate": 4.446089217843569e-06, "loss": 0.0114, "step": 13893 }, { "epoch": 4.31, "learning_rate": 4.445689137827566e-06, "loss": 0.0133, "step": 13894 }, { "epoch": 4.31, "learning_rate": 4.445289057811563e-06, "loss": 0.0325, "step": 13895 }, { "epoch": 4.31, "learning_rate": 4.444888977795559e-06, "loss": 0.0098, "step": 13896 }, { "epoch": 4.31, "learning_rate": 4.444488897779556e-06, "loss": 0.0548, "step": 13897 }, { "epoch": 4.31, "learning_rate": 4.4440888177635535e-06, "loss": 0.0327, "step": 13898 }, { "epoch": 4.31, "learning_rate": 4.443688737747549e-06, "loss": 0.0192, "step": 13899 }, { "epoch": 4.31, "learning_rate": 4.443288657731547e-06, "loss": 0.0316, "step": 13900 }, { "epoch": 4.31, "learning_rate": 4.442888577715544e-06, "loss": 0.0153, "step": 13901 }, { "epoch": 4.31, "learning_rate": 4.44248849769954e-06, "loss": 0.0141, "step": 13902 }, { "epoch": 4.32, "learning_rate": 4.442088417683537e-06, "loss": 0.016, "step": 13903 }, { "epoch": 4.32, "learning_rate": 4.441688337667534e-06, "loss": 0.0208, "step": 13904 }, { "epoch": 4.32, "learning_rate": 4.441288257651531e-06, "loss": 0.0233, "step": 13905 }, { "epoch": 4.32, "learning_rate": 4.440888177635527e-06, "loss": 0.0104, "step": 13906 }, { "epoch": 4.32, "learning_rate": 4.4404880976195246e-06, "loss": 0.0207, "step": 13907 }, { "epoch": 4.32, "learning_rate": 4.440088017603521e-06, "loss": 0.0197, "step": 13908 }, { "epoch": 4.32, "learning_rate": 4.439687937587518e-06, "loss": 0.0582, "step": 13909 }, { "epoch": 4.32, "learning_rate": 4.439287857571515e-06, "loss": 0.0219, "step": 13910 }, { "epoch": 4.32, "learning_rate": 4.438887777555511e-06, "loss": 0.0259, "step": 13911 }, { "epoch": 4.32, "learning_rate": 4.438487697539508e-06, "loss": 0.0187, "step": 13912 }, { "epoch": 4.32, "learning_rate": 4.438087617523505e-06, "loss": 0.0165, "step": 13913 }, { "epoch": 4.32, "learning_rate": 4.437687537507502e-06, "loss": 0.0302, "step": 13914 }, { "epoch": 4.32, "learning_rate": 4.437287457491499e-06, "loss": 0.016, "step": 13915 }, { "epoch": 4.32, "learning_rate": 4.436887377475496e-06, "loss": 0.0216, "step": 13916 }, { "epoch": 4.32, "learning_rate": 4.436487297459492e-06, "loss": 0.0081, "step": 13917 }, { "epoch": 4.32, "learning_rate": 4.436087217443489e-06, "loss": 0.0252, "step": 13918 }, { "epoch": 4.32, "learning_rate": 4.435687137427486e-06, "loss": 0.02, "step": 13919 }, { "epoch": 4.32, "learning_rate": 4.435287057411482e-06, "loss": 0.0057, "step": 13920 }, { "epoch": 4.32, "learning_rate": 4.43488697739548e-06, "loss": 0.0217, "step": 13921 }, { "epoch": 4.32, "learning_rate": 4.4344868973794766e-06, "loss": 0.0269, "step": 13922 }, { "epoch": 4.32, "learning_rate": 4.434086817363473e-06, "loss": 0.0257, "step": 13923 }, { "epoch": 4.32, "learning_rate": 4.43368673734747e-06, "loss": 0.04, "step": 13924 }, { "epoch": 4.32, "learning_rate": 4.433286657331467e-06, "loss": 0.0716, "step": 13925 }, { "epoch": 4.32, "learning_rate": 4.432886577315463e-06, "loss": 0.0155, "step": 13926 }, { "epoch": 4.32, "learning_rate": 4.43248649729946e-06, "loss": 0.0079, "step": 13927 }, { "epoch": 4.32, "learning_rate": 4.4320864172834575e-06, "loss": 0.0395, "step": 13928 }, { "epoch": 4.32, "learning_rate": 4.431686337267454e-06, "loss": 0.0174, "step": 13929 }, { "epoch": 4.32, "learning_rate": 4.431286257251451e-06, "loss": 0.1068, "step": 13930 }, { "epoch": 4.32, "learning_rate": 4.4308861772354476e-06, "loss": 0.0147, "step": 13931 }, { "epoch": 4.32, "learning_rate": 4.430486097219444e-06, "loss": 0.0794, "step": 13932 }, { "epoch": 4.32, "learning_rate": 4.430086017203441e-06, "loss": 0.1016, "step": 13933 }, { "epoch": 4.32, "learning_rate": 4.429685937187438e-06, "loss": 0.0186, "step": 13934 }, { "epoch": 4.33, "learning_rate": 4.429285857171435e-06, "loss": 0.0053, "step": 13935 }, { "epoch": 4.33, "learning_rate": 4.428885777155432e-06, "loss": 0.012, "step": 13936 }, { "epoch": 4.33, "learning_rate": 4.428485697139428e-06, "loss": 0.0735, "step": 13937 }, { "epoch": 4.33, "learning_rate": 4.428085617123425e-06, "loss": 0.0205, "step": 13938 }, { "epoch": 4.33, "learning_rate": 4.427685537107422e-06, "loss": 0.0187, "step": 13939 }, { "epoch": 4.33, "learning_rate": 4.427285457091419e-06, "loss": 0.0117, "step": 13940 }, { "epoch": 4.33, "learning_rate": 4.426885377075415e-06, "loss": 0.0235, "step": 13941 }, { "epoch": 4.33, "learning_rate": 4.426485297059413e-06, "loss": 0.0201, "step": 13942 }, { "epoch": 4.33, "learning_rate": 4.4260852170434095e-06, "loss": 0.0136, "step": 13943 }, { "epoch": 4.33, "learning_rate": 4.425685137027405e-06, "loss": 0.0103, "step": 13944 }, { "epoch": 4.33, "learning_rate": 4.425285057011403e-06, "loss": 0.0556, "step": 13945 }, { "epoch": 4.33, "learning_rate": 4.4248849769953996e-06, "loss": 0.0245, "step": 13946 }, { "epoch": 4.33, "learning_rate": 4.424484896979396e-06, "loss": 0.0174, "step": 13947 }, { "epoch": 4.33, "learning_rate": 4.424084816963393e-06, "loss": 0.0388, "step": 13948 }, { "epoch": 4.33, "learning_rate": 4.4236847369473905e-06, "loss": 0.0232, "step": 13949 }, { "epoch": 4.33, "learning_rate": 4.423284656931386e-06, "loss": 0.0256, "step": 13950 }, { "epoch": 4.33, "learning_rate": 4.422884576915383e-06, "loss": 0.017, "step": 13951 }, { "epoch": 4.33, "learning_rate": 4.4224844968993805e-06, "loss": 0.004, "step": 13952 }, { "epoch": 4.33, "learning_rate": 4.422084416883377e-06, "loss": 0.0187, "step": 13953 }, { "epoch": 4.33, "learning_rate": 4.421684336867374e-06, "loss": 0.0222, "step": 13954 }, { "epoch": 4.33, "learning_rate": 4.421284256851371e-06, "loss": 0.0128, "step": 13955 }, { "epoch": 4.33, "learning_rate": 4.420884176835367e-06, "loss": 0.0069, "step": 13956 }, { "epoch": 4.33, "learning_rate": 4.420484096819364e-06, "loss": 0.1003, "step": 13957 }, { "epoch": 4.33, "learning_rate": 4.420084016803361e-06, "loss": 0.009, "step": 13958 }, { "epoch": 4.33, "learning_rate": 4.419683936787358e-06, "loss": 0.0335, "step": 13959 }, { "epoch": 4.33, "learning_rate": 4.419283856771355e-06, "loss": 0.0107, "step": 13960 }, { "epoch": 4.33, "learning_rate": 4.4188837767553515e-06, "loss": 0.0175, "step": 13961 }, { "epoch": 4.33, "learning_rate": 4.418483696739348e-06, "loss": 0.0274, "step": 13962 }, { "epoch": 4.33, "learning_rate": 4.418083616723345e-06, "loss": 0.0196, "step": 13963 }, { "epoch": 4.33, "learning_rate": 4.417683536707342e-06, "loss": 0.0395, "step": 13964 }, { "epoch": 4.33, "learning_rate": 4.417283456691338e-06, "loss": 0.0265, "step": 13965 }, { "epoch": 4.33, "learning_rate": 4.416883376675336e-06, "loss": 0.0198, "step": 13966 }, { "epoch": 4.34, "learning_rate": 4.4164832966593325e-06, "loss": 0.0208, "step": 13967 }, { "epoch": 4.34, "learning_rate": 4.416083216643329e-06, "loss": 0.0159, "step": 13968 }, { "epoch": 4.34, "learning_rate": 4.415683136627326e-06, "loss": 0.0167, "step": 13969 }, { "epoch": 4.34, "learning_rate": 4.4152830566113226e-06, "loss": 0.01, "step": 13970 }, { "epoch": 4.34, "learning_rate": 4.414882976595319e-06, "loss": 0.0202, "step": 13971 }, { "epoch": 4.34, "learning_rate": 4.414482896579316e-06, "loss": 0.0299, "step": 13972 }, { "epoch": 4.34, "learning_rate": 4.4140828165633135e-06, "loss": 0.0125, "step": 13973 }, { "epoch": 4.34, "learning_rate": 4.41368273654731e-06, "loss": 0.0058, "step": 13974 }, { "epoch": 4.34, "learning_rate": 4.413282656531306e-06, "loss": 0.0114, "step": 13975 }, { "epoch": 4.34, "learning_rate": 4.4128825765153035e-06, "loss": 0.0126, "step": 13976 }, { "epoch": 4.34, "learning_rate": 4.4124824964993e-06, "loss": 0.0109, "step": 13977 }, { "epoch": 4.34, "learning_rate": 4.412082416483297e-06, "loss": 0.0269, "step": 13978 }, { "epoch": 4.34, "learning_rate": 4.411682336467294e-06, "loss": 0.0092, "step": 13979 }, { "epoch": 4.34, "learning_rate": 4.411282256451291e-06, "loss": 0.0268, "step": 13980 }, { "epoch": 4.34, "learning_rate": 4.410882176435287e-06, "loss": 0.0174, "step": 13981 }, { "epoch": 4.34, "learning_rate": 4.410482096419284e-06, "loss": 0.0242, "step": 13982 }, { "epoch": 4.34, "learning_rate": 4.410082016403281e-06, "loss": 0.0201, "step": 13983 }, { "epoch": 4.34, "learning_rate": 4.409681936387278e-06, "loss": 0.0054, "step": 13984 }, { "epoch": 4.34, "learning_rate": 4.4092818563712746e-06, "loss": 0.0171, "step": 13985 }, { "epoch": 4.34, "learning_rate": 4.408881776355271e-06, "loss": 0.0127, "step": 13986 }, { "epoch": 4.34, "learning_rate": 4.408481696339269e-06, "loss": 0.0262, "step": 13987 }, { "epoch": 4.34, "learning_rate": 4.408081616323265e-06, "loss": 0.0456, "step": 13988 }, { "epoch": 4.34, "learning_rate": 4.407681536307261e-06, "loss": 0.0141, "step": 13989 }, { "epoch": 4.34, "learning_rate": 4.407281456291259e-06, "loss": 0.0038, "step": 13990 }, { "epoch": 4.34, "learning_rate": 4.4068813762752555e-06, "loss": 0.0099, "step": 13991 }, { "epoch": 4.34, "learning_rate": 4.406481296259252e-06, "loss": 0.0153, "step": 13992 }, { "epoch": 4.34, "learning_rate": 4.406081216243249e-06, "loss": 0.0129, "step": 13993 }, { "epoch": 4.34, "learning_rate": 4.405681136227246e-06, "loss": 0.0173, "step": 13994 }, { "epoch": 4.34, "learning_rate": 4.405281056211242e-06, "loss": 0.016, "step": 13995 }, { "epoch": 4.34, "learning_rate": 4.404880976195239e-06, "loss": 0.012, "step": 13996 }, { "epoch": 4.34, "learning_rate": 4.4044808961792365e-06, "loss": 0.0093, "step": 13997 }, { "epoch": 4.34, "learning_rate": 4.404080816163233e-06, "loss": 0.0159, "step": 13998 }, { "epoch": 4.34, "learning_rate": 4.40368073614723e-06, "loss": 0.0134, "step": 13999 }, { "epoch": 4.35, "learning_rate": 4.4032806561312265e-06, "loss": 0.0346, "step": 14000 }, { "epoch": 4.35, "learning_rate": 4.402880576115223e-06, "loss": 0.0121, "step": 14001 }, { "epoch": 4.35, "learning_rate": 4.40248049609922e-06, "loss": 0.0158, "step": 14002 }, { "epoch": 4.35, "learning_rate": 4.402080416083217e-06, "loss": 0.0186, "step": 14003 }, { "epoch": 4.35, "learning_rate": 4.401680336067214e-06, "loss": 0.0263, "step": 14004 }, { "epoch": 4.35, "learning_rate": 4.401680336067214e-06, "loss": 0.1601, "step": 14005 }, { "epoch": 4.35, "learning_rate": 4.401280256051211e-06, "loss": 0.032, "step": 14006 }, { "epoch": 4.35, "learning_rate": 4.4008801760352075e-06, "loss": 0.0256, "step": 14007 }, { "epoch": 4.35, "learning_rate": 4.400480096019204e-06, "loss": 0.0388, "step": 14008 }, { "epoch": 4.35, "learning_rate": 4.400080016003201e-06, "loss": 0.0425, "step": 14009 }, { "epoch": 4.35, "learning_rate": 4.3996799359871976e-06, "loss": 0.0186, "step": 14010 }, { "epoch": 4.35, "learning_rate": 4.399279855971194e-06, "loss": 0.0162, "step": 14011 }, { "epoch": 4.35, "learning_rate": 4.398879775955192e-06, "loss": 0.009, "step": 14012 }, { "epoch": 4.35, "learning_rate": 4.3984796959391885e-06, "loss": 0.0138, "step": 14013 }, { "epoch": 4.35, "learning_rate": 4.398079615923185e-06, "loss": 0.0071, "step": 14014 }, { "epoch": 4.35, "learning_rate": 4.397679535907182e-06, "loss": 0.0149, "step": 14015 }, { "epoch": 4.35, "learning_rate": 4.3972794558911785e-06, "loss": 0.0093, "step": 14016 }, { "epoch": 4.35, "learning_rate": 4.396879375875175e-06, "loss": 0.013, "step": 14017 }, { "epoch": 4.35, "learning_rate": 4.396479295859172e-06, "loss": 0.0237, "step": 14018 }, { "epoch": 4.35, "learning_rate": 4.3960792158431694e-06, "loss": 0.016, "step": 14019 }, { "epoch": 4.35, "learning_rate": 4.395679135827165e-06, "loss": 0.0086, "step": 14020 }, { "epoch": 4.35, "learning_rate": 4.395279055811163e-06, "loss": 0.0127, "step": 14021 }, { "epoch": 4.35, "learning_rate": 4.3948789757951595e-06, "loss": 0.0113, "step": 14022 }, { "epoch": 4.35, "learning_rate": 4.394478895779156e-06, "loss": 0.0329, "step": 14023 }, { "epoch": 4.35, "learning_rate": 4.394078815763153e-06, "loss": 0.0271, "step": 14024 }, { "epoch": 4.35, "learning_rate": 4.3936787357471496e-06, "loss": 0.0328, "step": 14025 }, { "epoch": 4.35, "learning_rate": 4.393278655731146e-06, "loss": 0.0298, "step": 14026 }, { "epoch": 4.35, "learning_rate": 4.392878575715143e-06, "loss": 0.0192, "step": 14027 }, { "epoch": 4.35, "learning_rate": 4.3924784956991405e-06, "loss": 0.0056, "step": 14028 }, { "epoch": 4.35, "learning_rate": 4.392078415683137e-06, "loss": 0.0315, "step": 14029 }, { "epoch": 4.35, "learning_rate": 4.391678335667134e-06, "loss": 0.0154, "step": 14030 }, { "epoch": 4.35, "learning_rate": 4.3912782556511305e-06, "loss": 0.0227, "step": 14031 }, { "epoch": 4.36, "learning_rate": 4.390878175635128e-06, "loss": 0.0807, "step": 14032 }, { "epoch": 4.36, "learning_rate": 4.390478095619124e-06, "loss": 0.0196, "step": 14033 }, { "epoch": 4.36, "learning_rate": 4.390078015603121e-06, "loss": 0.0094, "step": 14034 }, { "epoch": 4.36, "learning_rate": 4.389677935587118e-06, "loss": 0.0074, "step": 14035 }, { "epoch": 4.36, "learning_rate": 4.389277855571115e-06, "loss": 0.0191, "step": 14036 }, { "epoch": 4.36, "learning_rate": 4.3888777755551115e-06, "loss": 0.0304, "step": 14037 }, { "epoch": 4.36, "learning_rate": 4.388477695539108e-06, "loss": 0.0134, "step": 14038 }, { "epoch": 4.36, "learning_rate": 4.388077615523105e-06, "loss": 0.0248, "step": 14039 }, { "epoch": 4.36, "learning_rate": 4.3876775355071015e-06, "loss": 0.0055, "step": 14040 }, { "epoch": 4.36, "learning_rate": 4.387277455491098e-06, "loss": 0.0419, "step": 14041 }, { "epoch": 4.36, "learning_rate": 4.386877375475096e-06, "loss": 0.0065, "step": 14042 }, { "epoch": 4.36, "learning_rate": 4.3864772954590925e-06, "loss": 0.0064, "step": 14043 }, { "epoch": 4.36, "learning_rate": 4.386077215443089e-06, "loss": 0.0138, "step": 14044 }, { "epoch": 4.36, "learning_rate": 4.385677135427086e-06, "loss": 0.015, "step": 14045 }, { "epoch": 4.36, "learning_rate": 4.3852770554110825e-06, "loss": 0.0261, "step": 14046 }, { "epoch": 4.36, "learning_rate": 4.384876975395079e-06, "loss": 0.0225, "step": 14047 }, { "epoch": 4.36, "learning_rate": 4.384476895379076e-06, "loss": 0.0268, "step": 14048 }, { "epoch": 4.36, "learning_rate": 4.384076815363073e-06, "loss": 0.0207, "step": 14049 }, { "epoch": 4.36, "learning_rate": 4.38367673534707e-06, "loss": 0.0104, "step": 14050 }, { "epoch": 4.36, "learning_rate": 4.383276655331067e-06, "loss": 0.0881, "step": 14051 }, { "epoch": 4.36, "learning_rate": 4.3828765753150635e-06, "loss": 0.0192, "step": 14052 }, { "epoch": 4.36, "learning_rate": 4.38247649529906e-06, "loss": 0.0293, "step": 14053 }, { "epoch": 4.36, "learning_rate": 4.382076415283057e-06, "loss": 0.0222, "step": 14054 }, { "epoch": 4.36, "learning_rate": 4.3816763352670535e-06, "loss": 0.0481, "step": 14055 }, { "epoch": 4.36, "learning_rate": 4.381276255251051e-06, "loss": 0.0226, "step": 14056 }, { "epoch": 4.36, "learning_rate": 4.380876175235048e-06, "loss": 0.0199, "step": 14057 }, { "epoch": 4.36, "learning_rate": 4.380476095219044e-06, "loss": 0.0081, "step": 14058 }, { "epoch": 4.36, "learning_rate": 4.380076015203041e-06, "loss": 0.0133, "step": 14059 }, { "epoch": 4.36, "learning_rate": 4.379675935187038e-06, "loss": 0.0186, "step": 14060 }, { "epoch": 4.36, "learning_rate": 4.3792758551710345e-06, "loss": 0.0164, "step": 14061 }, { "epoch": 4.36, "learning_rate": 4.378875775155031e-06, "loss": 0.0103, "step": 14062 }, { "epoch": 4.36, "learning_rate": 4.378475695139029e-06, "loss": 0.008, "step": 14063 }, { "epoch": 4.37, "learning_rate": 4.3780756151230246e-06, "loss": 0.0108, "step": 14064 }, { "epoch": 4.37, "learning_rate": 4.377675535107021e-06, "loss": 0.0483, "step": 14065 }, { "epoch": 4.37, "learning_rate": 4.377275455091019e-06, "loss": 0.0144, "step": 14066 }, { "epoch": 4.37, "learning_rate": 4.3768753750750155e-06, "loss": 0.0113, "step": 14067 }, { "epoch": 4.37, "learning_rate": 4.376475295059012e-06, "loss": 0.0268, "step": 14068 }, { "epoch": 4.37, "learning_rate": 4.376075215043009e-06, "loss": 0.0176, "step": 14069 }, { "epoch": 4.37, "learning_rate": 4.375675135027006e-06, "loss": 0.0139, "step": 14070 }, { "epoch": 4.37, "learning_rate": 4.375275055011002e-06, "loss": 0.0093, "step": 14071 }, { "epoch": 4.37, "learning_rate": 4.374874974994999e-06, "loss": 0.0103, "step": 14072 }, { "epoch": 4.37, "learning_rate": 4.3744748949789964e-06, "loss": 0.0157, "step": 14073 }, { "epoch": 4.37, "learning_rate": 4.374074814962993e-06, "loss": 0.0763, "step": 14074 }, { "epoch": 4.37, "learning_rate": 4.37367473494699e-06, "loss": 0.0156, "step": 14075 }, { "epoch": 4.37, "learning_rate": 4.3732746549309865e-06, "loss": 0.0248, "step": 14076 }, { "epoch": 4.37, "learning_rate": 4.372874574914983e-06, "loss": 0.1213, "step": 14077 }, { "epoch": 4.37, "learning_rate": 4.37247449489898e-06, "loss": 0.0246, "step": 14078 }, { "epoch": 4.37, "learning_rate": 4.3720744148829765e-06, "loss": 0.0151, "step": 14079 }, { "epoch": 4.37, "learning_rate": 4.371674334866974e-06, "loss": 0.0131, "step": 14080 }, { "epoch": 4.37, "learning_rate": 4.371274254850971e-06, "loss": 0.0816, "step": 14081 }, { "epoch": 4.37, "learning_rate": 4.3708741748349675e-06, "loss": 0.0231, "step": 14082 }, { "epoch": 4.37, "learning_rate": 4.370474094818964e-06, "loss": 0.0337, "step": 14083 }, { "epoch": 4.37, "learning_rate": 4.370074014802961e-06, "loss": 0.0234, "step": 14084 }, { "epoch": 4.37, "learning_rate": 4.3696739347869575e-06, "loss": 0.0797, "step": 14085 }, { "epoch": 4.37, "learning_rate": 4.369273854770954e-06, "loss": 0.0061, "step": 14086 }, { "epoch": 4.37, "learning_rate": 4.368873774754952e-06, "loss": 0.0136, "step": 14087 }, { "epoch": 4.37, "learning_rate": 4.368473694738948e-06, "loss": 0.0055, "step": 14088 }, { "epoch": 4.37, "learning_rate": 4.368073614722945e-06, "loss": 0.0137, "step": 14089 }, { "epoch": 4.37, "learning_rate": 4.367673534706942e-06, "loss": 0.0798, "step": 14090 }, { "epoch": 4.37, "learning_rate": 4.3672734546909385e-06, "loss": 0.0112, "step": 14091 }, { "epoch": 4.37, "learning_rate": 4.366873374674935e-06, "loss": 0.048, "step": 14092 }, { "epoch": 4.37, "learning_rate": 4.366473294658932e-06, "loss": 0.0496, "step": 14093 }, { "epoch": 4.37, "learning_rate": 4.366073214642929e-06, "loss": 0.0314, "step": 14094 }, { "epoch": 4.37, "learning_rate": 4.365673134626926e-06, "loss": 0.0086, "step": 14095 }, { "epoch": 4.38, "learning_rate": 4.365273054610922e-06, "loss": 0.009, "step": 14096 }, { "epoch": 4.38, "learning_rate": 4.3648729745949194e-06, "loss": 0.0334, "step": 14097 }, { "epoch": 4.38, "learning_rate": 4.364472894578916e-06, "loss": 0.0163, "step": 14098 }, { "epoch": 4.38, "learning_rate": 4.364072814562913e-06, "loss": 0.0113, "step": 14099 }, { "epoch": 4.38, "learning_rate": 4.3636727345469095e-06, "loss": 0.0921, "step": 14100 }, { "epoch": 4.38, "learning_rate": 4.363272654530907e-06, "loss": 0.0068, "step": 14101 }, { "epoch": 4.38, "learning_rate": 4.362872574514903e-06, "loss": 0.0091, "step": 14102 }, { "epoch": 4.38, "learning_rate": 4.3624724944988996e-06, "loss": 0.0228, "step": 14103 }, { "epoch": 4.38, "learning_rate": 4.362072414482897e-06, "loss": 0.0476, "step": 14104 }, { "epoch": 4.38, "learning_rate": 4.361672334466894e-06, "loss": 0.0073, "step": 14105 }, { "epoch": 4.38, "learning_rate": 4.3612722544508905e-06, "loss": 0.0139, "step": 14106 }, { "epoch": 4.38, "learning_rate": 4.360872174434887e-06, "loss": 0.0139, "step": 14107 }, { "epoch": 4.38, "learning_rate": 4.360472094418884e-06, "loss": 0.0104, "step": 14108 }, { "epoch": 4.38, "learning_rate": 4.3600720144028805e-06, "loss": 0.0203, "step": 14109 }, { "epoch": 4.38, "learning_rate": 4.359671934386877e-06, "loss": 0.0171, "step": 14110 }, { "epoch": 4.38, "learning_rate": 4.359271854370875e-06, "loss": 0.0407, "step": 14111 }, { "epoch": 4.38, "learning_rate": 4.3588717743548714e-06, "loss": 0.0169, "step": 14112 }, { "epoch": 4.38, "learning_rate": 4.358471694338868e-06, "loss": 0.0189, "step": 14113 }, { "epoch": 4.38, "learning_rate": 4.358071614322865e-06, "loss": 0.0031, "step": 14114 }, { "epoch": 4.38, "learning_rate": 4.3576715343068615e-06, "loss": 0.0143, "step": 14115 }, { "epoch": 4.38, "learning_rate": 4.357271454290858e-06, "loss": 0.0243, "step": 14116 }, { "epoch": 4.38, "learning_rate": 4.356871374274856e-06, "loss": 0.0151, "step": 14117 }, { "epoch": 4.38, "learning_rate": 4.356471294258852e-06, "loss": 0.044, "step": 14118 }, { "epoch": 4.38, "learning_rate": 4.356071214242849e-06, "loss": 0.0102, "step": 14119 }, { "epoch": 4.38, "learning_rate": 4.355671134226846e-06, "loss": 0.0242, "step": 14120 }, { "epoch": 4.38, "learning_rate": 4.3552710542108425e-06, "loss": 0.0192, "step": 14121 }, { "epoch": 4.38, "learning_rate": 4.354870974194839e-06, "loss": 0.0361, "step": 14122 }, { "epoch": 4.38, "learning_rate": 4.354470894178836e-06, "loss": 0.0128, "step": 14123 }, { "epoch": 4.38, "learning_rate": 4.354070814162833e-06, "loss": 0.0233, "step": 14124 }, { "epoch": 4.38, "learning_rate": 4.35367073414683e-06, "loss": 0.0294, "step": 14125 }, { "epoch": 4.38, "learning_rate": 4.353270654130827e-06, "loss": 0.0033, "step": 14126 }, { "epoch": 4.38, "learning_rate": 4.352870574114823e-06, "loss": 0.0548, "step": 14127 }, { "epoch": 4.39, "learning_rate": 4.35247049409882e-06, "loss": 0.0051, "step": 14128 }, { "epoch": 4.39, "learning_rate": 4.352070414082817e-06, "loss": 0.026, "step": 14129 }, { "epoch": 4.39, "learning_rate": 4.3516703340668135e-06, "loss": 0.014, "step": 14130 }, { "epoch": 4.39, "learning_rate": 4.351270254050811e-06, "loss": 0.0083, "step": 14131 }, { "epoch": 4.39, "learning_rate": 4.350870174034808e-06, "loss": 0.0088, "step": 14132 }, { "epoch": 4.39, "learning_rate": 4.350470094018804e-06, "loss": 0.0372, "step": 14133 }, { "epoch": 4.39, "learning_rate": 4.350070014002801e-06, "loss": 0.015, "step": 14134 }, { "epoch": 4.39, "learning_rate": 4.349669933986798e-06, "loss": 0.0933, "step": 14135 }, { "epoch": 4.39, "learning_rate": 4.3492698539707944e-06, "loss": 0.0432, "step": 14136 }, { "epoch": 4.39, "learning_rate": 4.348869773954791e-06, "loss": 0.0203, "step": 14137 }, { "epoch": 4.39, "learning_rate": 4.348469693938789e-06, "loss": 0.0083, "step": 14138 }, { "epoch": 4.39, "learning_rate": 4.348069613922785e-06, "loss": 0.0274, "step": 14139 }, { "epoch": 4.39, "learning_rate": 4.347669533906781e-06, "loss": 0.0255, "step": 14140 }, { "epoch": 4.39, "learning_rate": 4.347269453890779e-06, "loss": 0.0156, "step": 14141 }, { "epoch": 4.39, "learning_rate": 4.346869373874775e-06, "loss": 0.0094, "step": 14142 }, { "epoch": 4.39, "learning_rate": 4.346469293858772e-06, "loss": 0.0027, "step": 14143 }, { "epoch": 4.39, "learning_rate": 4.346069213842769e-06, "loss": 0.008, "step": 14144 }, { "epoch": 4.39, "learning_rate": 4.345669133826766e-06, "loss": 0.0178, "step": 14145 }, { "epoch": 4.39, "learning_rate": 4.345269053810762e-06, "loss": 0.0211, "step": 14146 }, { "epoch": 4.39, "learning_rate": 4.344868973794759e-06, "loss": 0.0088, "step": 14147 }, { "epoch": 4.39, "learning_rate": 4.344468893778756e-06, "loss": 0.0241, "step": 14148 }, { "epoch": 4.39, "learning_rate": 4.344068813762753e-06, "loss": 0.013, "step": 14149 }, { "epoch": 4.39, "learning_rate": 4.34366873374675e-06, "loss": 0.0079, "step": 14150 }, { "epoch": 4.39, "learning_rate": 4.3432686537307464e-06, "loss": 0.0139, "step": 14151 }, { "epoch": 4.39, "learning_rate": 4.342868573714744e-06, "loss": 0.0271, "step": 14152 }, { "epoch": 4.39, "learning_rate": 4.34246849369874e-06, "loss": 0.015, "step": 14153 }, { "epoch": 4.39, "learning_rate": 4.3420684136827365e-06, "loss": 0.0688, "step": 14154 }, { "epoch": 4.39, "learning_rate": 4.341668333666734e-06, "loss": 0.0436, "step": 14155 }, { "epoch": 4.39, "learning_rate": 4.341268253650731e-06, "loss": 0.0078, "step": 14156 }, { "epoch": 4.39, "learning_rate": 4.340868173634727e-06, "loss": 0.0285, "step": 14157 }, { "epoch": 4.39, "learning_rate": 4.340468093618724e-06, "loss": 0.0185, "step": 14158 }, { "epoch": 4.39, "learning_rate": 4.340068013602721e-06, "loss": 0.0392, "step": 14159 }, { "epoch": 4.39, "learning_rate": 4.3396679335867175e-06, "loss": 0.0206, "step": 14160 }, { "epoch": 4.4, "learning_rate": 4.339267853570714e-06, "loss": 0.0142, "step": 14161 }, { "epoch": 4.4, "learning_rate": 4.338867773554712e-06, "loss": 0.0386, "step": 14162 }, { "epoch": 4.4, "learning_rate": 4.338467693538708e-06, "loss": 0.0086, "step": 14163 }, { "epoch": 4.4, "learning_rate": 4.338067613522705e-06, "loss": 0.0105, "step": 14164 }, { "epoch": 4.4, "learning_rate": 4.337667533506702e-06, "loss": 0.0082, "step": 14165 }, { "epoch": 4.4, "learning_rate": 4.337267453490698e-06, "loss": 0.0092, "step": 14166 }, { "epoch": 4.4, "learning_rate": 4.336867373474695e-06, "loss": 0.1005, "step": 14167 }, { "epoch": 4.4, "learning_rate": 4.336467293458692e-06, "loss": 0.0246, "step": 14168 }, { "epoch": 4.4, "learning_rate": 4.336067213442689e-06, "loss": 0.0055, "step": 14169 }, { "epoch": 4.4, "learning_rate": 4.335667133426686e-06, "loss": 0.0196, "step": 14170 }, { "epoch": 4.4, "learning_rate": 4.335267053410682e-06, "loss": 0.0914, "step": 14171 }, { "epoch": 4.4, "learning_rate": 4.334866973394679e-06, "loss": 0.0058, "step": 14172 }, { "epoch": 4.4, "learning_rate": 4.334466893378676e-06, "loss": 0.018, "step": 14173 }, { "epoch": 4.4, "learning_rate": 4.334066813362673e-06, "loss": 0.0278, "step": 14174 }, { "epoch": 4.4, "learning_rate": 4.3336667333466694e-06, "loss": 0.0339, "step": 14175 }, { "epoch": 4.4, "learning_rate": 4.333266653330667e-06, "loss": 0.0107, "step": 14176 }, { "epoch": 4.4, "learning_rate": 4.332866573314664e-06, "loss": 0.011, "step": 14177 }, { "epoch": 4.4, "learning_rate": 4.3324664932986595e-06, "loss": 0.01, "step": 14178 }, { "epoch": 4.4, "learning_rate": 4.332066413282657e-06, "loss": 0.0185, "step": 14179 }, { "epoch": 4.4, "learning_rate": 4.331666333266654e-06, "loss": 0.0107, "step": 14180 }, { "epoch": 4.4, "learning_rate": 4.33126625325065e-06, "loss": 0.0032, "step": 14181 }, { "epoch": 4.4, "learning_rate": 4.330866173234647e-06, "loss": 0.0301, "step": 14182 }, { "epoch": 4.4, "learning_rate": 4.330466093218645e-06, "loss": 0.0151, "step": 14183 }, { "epoch": 4.4, "learning_rate": 4.3300660132026405e-06, "loss": 0.0213, "step": 14184 }, { "epoch": 4.4, "learning_rate": 4.329665933186637e-06, "loss": 0.0279, "step": 14185 }, { "epoch": 4.4, "learning_rate": 4.329265853170635e-06, "loss": 0.0108, "step": 14186 }, { "epoch": 4.4, "learning_rate": 4.328865773154631e-06, "loss": 0.0955, "step": 14187 }, { "epoch": 4.4, "learning_rate": 4.328465693138628e-06, "loss": 0.0083, "step": 14188 }, { "epoch": 4.4, "learning_rate": 4.328065613122625e-06, "loss": 0.0106, "step": 14189 }, { "epoch": 4.4, "learning_rate": 4.3276655331066214e-06, "loss": 0.0305, "step": 14190 }, { "epoch": 4.4, "learning_rate": 4.327265453090618e-06, "loss": 0.0098, "step": 14191 }, { "epoch": 4.4, "learning_rate": 4.326865373074615e-06, "loss": 0.0436, "step": 14192 }, { "epoch": 4.41, "learning_rate": 4.326465293058612e-06, "loss": 0.0099, "step": 14193 }, { "epoch": 4.41, "learning_rate": 4.326065213042609e-06, "loss": 0.0099, "step": 14194 }, { "epoch": 4.41, "learning_rate": 4.325665133026606e-06, "loss": 0.0099, "step": 14195 }, { "epoch": 4.41, "learning_rate": 4.325265053010602e-06, "loss": 0.0344, "step": 14196 }, { "epoch": 4.41, "learning_rate": 4.324864972994599e-06, "loss": 0.0136, "step": 14197 }, { "epoch": 4.41, "learning_rate": 4.324464892978596e-06, "loss": 0.0234, "step": 14198 }, { "epoch": 4.41, "learning_rate": 4.3240648129625925e-06, "loss": 0.0529, "step": 14199 }, { "epoch": 4.41, "learning_rate": 4.32366473294659e-06, "loss": 0.0164, "step": 14200 }, { "epoch": 4.41, "learning_rate": 4.323264652930587e-06, "loss": 0.0131, "step": 14201 }, { "epoch": 4.41, "learning_rate": 4.322864572914583e-06, "loss": 0.0099, "step": 14202 }, { "epoch": 4.41, "learning_rate": 4.32246449289858e-06, "loss": 0.0217, "step": 14203 }, { "epoch": 4.41, "learning_rate": 4.322064412882577e-06, "loss": 0.0333, "step": 14204 }, { "epoch": 4.41, "learning_rate": 4.321664332866573e-06, "loss": 0.0136, "step": 14205 }, { "epoch": 4.41, "learning_rate": 4.32126425285057e-06, "loss": 0.0477, "step": 14206 }, { "epoch": 4.41, "learning_rate": 4.320864172834568e-06, "loss": 0.015, "step": 14207 }, { "epoch": 4.41, "learning_rate": 4.320464092818564e-06, "loss": 0.0201, "step": 14208 }, { "epoch": 4.41, "learning_rate": 4.320064012802561e-06, "loss": 0.0098, "step": 14209 }, { "epoch": 4.41, "learning_rate": 4.319663932786558e-06, "loss": 0.0049, "step": 14210 }, { "epoch": 4.41, "learning_rate": 4.319263852770554e-06, "loss": 0.0068, "step": 14211 }, { "epoch": 4.41, "learning_rate": 4.318863772754551e-06, "loss": 0.0231, "step": 14212 }, { "epoch": 4.41, "learning_rate": 4.318463692738548e-06, "loss": 0.0148, "step": 14213 }, { "epoch": 4.41, "learning_rate": 4.318063612722545e-06, "loss": 0.0115, "step": 14214 }, { "epoch": 4.41, "learning_rate": 4.317663532706542e-06, "loss": 0.0259, "step": 14215 }, { "epoch": 4.41, "learning_rate": 4.317263452690539e-06, "loss": 0.0532, "step": 14216 }, { "epoch": 4.41, "learning_rate": 4.316863372674535e-06, "loss": 0.0148, "step": 14217 }, { "epoch": 4.41, "learning_rate": 4.316463292658532e-06, "loss": 0.0262, "step": 14218 }, { "epoch": 4.41, "learning_rate": 4.316063212642529e-06, "loss": 0.0189, "step": 14219 }, { "epoch": 4.41, "learning_rate": 4.315663132626525e-06, "loss": 0.0121, "step": 14220 }, { "epoch": 4.41, "learning_rate": 4.315263052610523e-06, "loss": 0.0139, "step": 14221 }, { "epoch": 4.41, "learning_rate": 4.314862972594519e-06, "loss": 0.0151, "step": 14222 }, { "epoch": 4.41, "learning_rate": 4.314462892578516e-06, "loss": 0.0106, "step": 14223 }, { "epoch": 4.41, "learning_rate": 4.314062812562513e-06, "loss": 0.0274, "step": 14224 }, { "epoch": 4.42, "learning_rate": 4.31366273254651e-06, "loss": 0.0103, "step": 14225 }, { "epoch": 4.42, "learning_rate": 4.313262652530506e-06, "loss": 0.0361, "step": 14226 }, { "epoch": 4.42, "learning_rate": 4.312862572514503e-06, "loss": 0.0138, "step": 14227 }, { "epoch": 4.42, "learning_rate": 4.3124624924985e-06, "loss": 0.0279, "step": 14228 }, { "epoch": 4.42, "learning_rate": 4.3120624124824964e-06, "loss": 0.0402, "step": 14229 }, { "epoch": 4.42, "learning_rate": 4.311662332466494e-06, "loss": 0.0359, "step": 14230 }, { "epoch": 4.42, "learning_rate": 4.311262252450491e-06, "loss": 0.0161, "step": 14231 }, { "epoch": 4.42, "learning_rate": 4.310862172434487e-06, "loss": 0.0343, "step": 14232 }, { "epoch": 4.42, "learning_rate": 4.310462092418484e-06, "loss": 0.0176, "step": 14233 }, { "epoch": 4.42, "learning_rate": 4.310062012402481e-06, "loss": 0.0054, "step": 14234 }, { "epoch": 4.42, "learning_rate": 4.309661932386477e-06, "loss": 0.0142, "step": 14235 }, { "epoch": 4.42, "learning_rate": 4.309261852370474e-06, "loss": 0.0377, "step": 14236 }, { "epoch": 4.42, "learning_rate": 4.308861772354472e-06, "loss": 0.0268, "step": 14237 }, { "epoch": 4.42, "learning_rate": 4.308461692338468e-06, "loss": 0.0758, "step": 14238 }, { "epoch": 4.42, "learning_rate": 4.308061612322465e-06, "loss": 0.0229, "step": 14239 }, { "epoch": 4.42, "learning_rate": 4.307661532306462e-06, "loss": 0.0146, "step": 14240 }, { "epoch": 4.42, "learning_rate": 4.307261452290458e-06, "loss": 0.0105, "step": 14241 }, { "epoch": 4.42, "learning_rate": 4.306861372274455e-06, "loss": 0.014, "step": 14242 }, { "epoch": 4.42, "learning_rate": 4.306461292258452e-06, "loss": 0.0161, "step": 14243 }, { "epoch": 4.42, "learning_rate": 4.306061212242449e-06, "loss": 0.0077, "step": 14244 }, { "epoch": 4.42, "learning_rate": 4.305661132226446e-06, "loss": 0.0072, "step": 14245 }, { "epoch": 4.42, "learning_rate": 4.305261052210443e-06, "loss": 0.035, "step": 14246 }, { "epoch": 4.42, "learning_rate": 4.304860972194439e-06, "loss": 0.0173, "step": 14247 }, { "epoch": 4.42, "learning_rate": 4.304460892178436e-06, "loss": 0.0124, "step": 14248 }, { "epoch": 4.42, "learning_rate": 4.304060812162433e-06, "loss": 0.0355, "step": 14249 }, { "epoch": 4.42, "learning_rate": 4.303660732146429e-06, "loss": 0.012, "step": 14250 }, { "epoch": 4.42, "learning_rate": 4.303260652130427e-06, "loss": 0.0173, "step": 14251 }, { "epoch": 4.42, "learning_rate": 4.302860572114424e-06, "loss": 0.0221, "step": 14252 }, { "epoch": 4.42, "learning_rate": 4.3024604920984194e-06, "loss": 0.0218, "step": 14253 }, { "epoch": 4.42, "learning_rate": 4.302060412082417e-06, "loss": 0.0236, "step": 14254 }, { "epoch": 4.42, "learning_rate": 4.301660332066414e-06, "loss": 0.0187, "step": 14255 }, { "epoch": 4.42, "learning_rate": 4.30126025205041e-06, "loss": 0.0271, "step": 14256 }, { "epoch": 4.43, "learning_rate": 4.300860172034407e-06, "loss": 0.0737, "step": 14257 }, { "epoch": 4.43, "learning_rate": 4.3004600920184046e-06, "loss": 0.0104, "step": 14258 }, { "epoch": 4.43, "learning_rate": 4.300060012002401e-06, "loss": 0.006, "step": 14259 }, { "epoch": 4.43, "learning_rate": 4.299659931986397e-06, "loss": 0.0228, "step": 14260 }, { "epoch": 4.43, "learning_rate": 4.299259851970395e-06, "loss": 0.0151, "step": 14261 }, { "epoch": 4.43, "learning_rate": 4.298859771954391e-06, "loss": 0.0158, "step": 14262 }, { "epoch": 4.43, "learning_rate": 4.298459691938388e-06, "loss": 0.0551, "step": 14263 }, { "epoch": 4.43, "learning_rate": 4.298059611922385e-06, "loss": 0.0265, "step": 14264 }, { "epoch": 4.43, "learning_rate": 4.297659531906382e-06, "loss": 0.0207, "step": 14265 }, { "epoch": 4.43, "learning_rate": 4.297259451890378e-06, "loss": 0.0086, "step": 14266 }, { "epoch": 4.43, "learning_rate": 4.296859371874375e-06, "loss": 0.0182, "step": 14267 }, { "epoch": 4.43, "learning_rate": 4.296459291858372e-06, "loss": 0.0153, "step": 14268 }, { "epoch": 4.43, "learning_rate": 4.296059211842369e-06, "loss": 0.0114, "step": 14269 }, { "epoch": 4.43, "learning_rate": 4.295659131826366e-06, "loss": 0.0255, "step": 14270 }, { "epoch": 4.43, "learning_rate": 4.295259051810362e-06, "loss": 0.0394, "step": 14271 }, { "epoch": 4.43, "learning_rate": 4.294858971794359e-06, "loss": 0.0168, "step": 14272 }, { "epoch": 4.43, "learning_rate": 4.294458891778356e-06, "loss": 0.0133, "step": 14273 }, { "epoch": 4.43, "learning_rate": 4.294058811762352e-06, "loss": 0.0138, "step": 14274 }, { "epoch": 4.43, "learning_rate": 4.29365873174635e-06, "loss": 0.0076, "step": 14275 }, { "epoch": 4.43, "learning_rate": 4.293258651730347e-06, "loss": 0.0264, "step": 14276 }, { "epoch": 4.43, "learning_rate": 4.292858571714343e-06, "loss": 0.0115, "step": 14277 }, { "epoch": 4.43, "learning_rate": 4.29245849169834e-06, "loss": 0.0503, "step": 14278 }, { "epoch": 4.43, "learning_rate": 4.292058411682337e-06, "loss": 0.0185, "step": 14279 }, { "epoch": 4.43, "learning_rate": 4.291658331666333e-06, "loss": 0.0161, "step": 14280 }, { "epoch": 4.43, "learning_rate": 4.29125825165033e-06, "loss": 0.0108, "step": 14281 }, { "epoch": 4.43, "learning_rate": 4.290858171634328e-06, "loss": 0.0108, "step": 14282 }, { "epoch": 4.43, "learning_rate": 4.290458091618324e-06, "loss": 0.0156, "step": 14283 }, { "epoch": 4.43, "learning_rate": 4.290058011602321e-06, "loss": 0.016, "step": 14284 }, { "epoch": 4.43, "learning_rate": 4.289657931586318e-06, "loss": 0.0575, "step": 14285 }, { "epoch": 4.43, "learning_rate": 4.289257851570314e-06, "loss": 0.0301, "step": 14286 }, { "epoch": 4.43, "learning_rate": 4.288857771554311e-06, "loss": 0.0313, "step": 14287 }, { "epoch": 4.43, "learning_rate": 4.288457691538308e-06, "loss": 0.0184, "step": 14288 }, { "epoch": 4.43, "learning_rate": 4.288057611522305e-06, "loss": 0.0201, "step": 14289 }, { "epoch": 4.44, "learning_rate": 4.287657531506302e-06, "loss": 0.0776, "step": 14290 }, { "epoch": 4.44, "learning_rate": 4.287257451490298e-06, "loss": 0.0215, "step": 14291 }, { "epoch": 4.44, "learning_rate": 4.286857371474295e-06, "loss": 0.0107, "step": 14292 }, { "epoch": 4.44, "learning_rate": 4.286457291458292e-06, "loss": 0.0334, "step": 14293 }, { "epoch": 4.44, "learning_rate": 4.286057211442289e-06, "loss": 0.0138, "step": 14294 }, { "epoch": 4.44, "learning_rate": 4.285657131426285e-06, "loss": 0.0274, "step": 14295 }, { "epoch": 4.44, "learning_rate": 4.285257051410283e-06, "loss": 0.0232, "step": 14296 }, { "epoch": 4.44, "learning_rate": 4.2848569713942796e-06, "loss": 0.0068, "step": 14297 }, { "epoch": 4.44, "learning_rate": 4.284456891378275e-06, "loss": 0.0063, "step": 14298 }, { "epoch": 4.44, "learning_rate": 4.284056811362273e-06, "loss": 0.0081, "step": 14299 }, { "epoch": 4.44, "learning_rate": 4.28365673134627e-06, "loss": 0.037, "step": 14300 }, { "epoch": 4.44, "learning_rate": 4.283256651330266e-06, "loss": 0.0166, "step": 14301 }, { "epoch": 4.44, "learning_rate": 4.282856571314263e-06, "loss": 0.0178, "step": 14302 }, { "epoch": 4.44, "learning_rate": 4.2824564912982605e-06, "loss": 0.0143, "step": 14303 }, { "epoch": 4.44, "learning_rate": 4.282056411282256e-06, "loss": 0.029, "step": 14304 }, { "epoch": 4.44, "learning_rate": 4.281656331266253e-06, "loss": 0.0225, "step": 14305 }, { "epoch": 4.44, "learning_rate": 4.281256251250251e-06, "loss": 0.0066, "step": 14306 }, { "epoch": 4.44, "learning_rate": 4.280856171234247e-06, "loss": 0.0186, "step": 14307 }, { "epoch": 4.44, "learning_rate": 4.280456091218244e-06, "loss": 0.0048, "step": 14308 }, { "epoch": 4.44, "learning_rate": 4.280056011202241e-06, "loss": 0.0152, "step": 14309 }, { "epoch": 4.44, "learning_rate": 4.279655931186237e-06, "loss": 0.013, "step": 14310 }, { "epoch": 4.44, "learning_rate": 4.279255851170234e-06, "loss": 0.0084, "step": 14311 }, { "epoch": 4.44, "learning_rate": 4.278855771154231e-06, "loss": 0.019, "step": 14312 }, { "epoch": 4.44, "learning_rate": 4.278455691138228e-06, "loss": 0.0147, "step": 14313 }, { "epoch": 4.44, "learning_rate": 4.278055611122225e-06, "loss": 0.0649, "step": 14314 }, { "epoch": 4.44, "learning_rate": 4.277655531106222e-06, "loss": 0.0203, "step": 14315 }, { "epoch": 4.44, "learning_rate": 4.277255451090218e-06, "loss": 0.0074, "step": 14316 }, { "epoch": 4.44, "learning_rate": 4.276855371074215e-06, "loss": 0.0119, "step": 14317 }, { "epoch": 4.44, "learning_rate": 4.276455291058212e-06, "loss": 0.0075, "step": 14318 }, { "epoch": 4.44, "learning_rate": 4.276055211042208e-06, "loss": 0.0183, "step": 14319 }, { "epoch": 4.44, "learning_rate": 4.275655131026206e-06, "loss": 0.0336, "step": 14320 }, { "epoch": 4.44, "learning_rate": 4.275255051010203e-06, "loss": 0.0157, "step": 14321 }, { "epoch": 4.45, "learning_rate": 4.274854970994199e-06, "loss": 0.0155, "step": 14322 }, { "epoch": 4.45, "learning_rate": 4.274454890978196e-06, "loss": 0.0327, "step": 14323 }, { "epoch": 4.45, "learning_rate": 4.274054810962193e-06, "loss": 0.0165, "step": 14324 }, { "epoch": 4.45, "learning_rate": 4.273654730946189e-06, "loss": 0.0059, "step": 14325 }, { "epoch": 4.45, "learning_rate": 4.273254650930186e-06, "loss": 0.0218, "step": 14326 }, { "epoch": 4.45, "learning_rate": 4.2728545709141835e-06, "loss": 0.0212, "step": 14327 }, { "epoch": 4.45, "learning_rate": 4.27245449089818e-06, "loss": 0.0106, "step": 14328 }, { "epoch": 4.45, "learning_rate": 4.272054410882177e-06, "loss": 0.0152, "step": 14329 }, { "epoch": 4.45, "learning_rate": 4.271654330866174e-06, "loss": 0.0152, "step": 14330 }, { "epoch": 4.45, "learning_rate": 4.27125425085017e-06, "loss": 0.0085, "step": 14331 }, { "epoch": 4.45, "learning_rate": 4.270854170834167e-06, "loss": 0.0086, "step": 14332 }, { "epoch": 4.45, "learning_rate": 4.2704540908181645e-06, "loss": 0.012, "step": 14333 }, { "epoch": 4.45, "learning_rate": 4.270054010802161e-06, "loss": 0.0155, "step": 14334 }, { "epoch": 4.45, "learning_rate": 4.269653930786157e-06, "loss": 0.0061, "step": 14335 }, { "epoch": 4.45, "learning_rate": 4.2692538507701546e-06, "loss": 0.051, "step": 14336 }, { "epoch": 4.45, "learning_rate": 4.268853770754151e-06, "loss": 0.0065, "step": 14337 }, { "epoch": 4.45, "learning_rate": 4.268453690738148e-06, "loss": 0.0263, "step": 14338 }, { "epoch": 4.45, "learning_rate": 4.268053610722145e-06, "loss": 0.043, "step": 14339 }, { "epoch": 4.45, "learning_rate": 4.267653530706142e-06, "loss": 0.0481, "step": 14340 }, { "epoch": 4.45, "learning_rate": 4.267253450690139e-06, "loss": 0.0191, "step": 14341 }, { "epoch": 4.45, "learning_rate": 4.266853370674135e-06, "loss": 0.0148, "step": 14342 }, { "epoch": 4.45, "learning_rate": 4.266453290658132e-06, "loss": 0.0376, "step": 14343 }, { "epoch": 4.45, "learning_rate": 4.266053210642129e-06, "loss": 0.0136, "step": 14344 }, { "epoch": 4.45, "learning_rate": 4.265653130626126e-06, "loss": 0.0472, "step": 14345 }, { "epoch": 4.45, "learning_rate": 4.265253050610122e-06, "loss": 0.0295, "step": 14346 }, { "epoch": 4.45, "learning_rate": 4.26485297059412e-06, "loss": 0.0743, "step": 14347 }, { "epoch": 4.45, "learning_rate": 4.264452890578116e-06, "loss": 0.0564, "step": 14348 }, { "epoch": 4.45, "learning_rate": 4.264052810562112e-06, "loss": 0.0261, "step": 14349 }, { "epoch": 4.45, "learning_rate": 4.26365273054611e-06, "loss": 0.0117, "step": 14350 }, { "epoch": 4.45, "learning_rate": 4.2632526505301066e-06, "loss": 0.0177, "step": 14351 }, { "epoch": 4.45, "learning_rate": 4.262852570514103e-06, "loss": 0.0067, "step": 14352 }, { "epoch": 4.45, "learning_rate": 4.2624524904981e-06, "loss": 0.0173, "step": 14353 }, { "epoch": 4.46, "learning_rate": 4.262052410482097e-06, "loss": 0.0152, "step": 14354 }, { "epoch": 4.46, "learning_rate": 4.261652330466093e-06, "loss": 0.014, "step": 14355 }, { "epoch": 4.46, "learning_rate": 4.26125225045009e-06, "loss": 0.0337, "step": 14356 }, { "epoch": 4.46, "learning_rate": 4.2608521704340875e-06, "loss": 0.0218, "step": 14357 }, { "epoch": 4.46, "learning_rate": 4.260452090418084e-06, "loss": 0.0749, "step": 14358 }, { "epoch": 4.46, "learning_rate": 4.260052010402081e-06, "loss": 0.098, "step": 14359 }, { "epoch": 4.46, "learning_rate": 4.259651930386078e-06, "loss": 0.0141, "step": 14360 }, { "epoch": 4.46, "learning_rate": 4.259251850370074e-06, "loss": 0.0182, "step": 14361 }, { "epoch": 4.46, "learning_rate": 4.258851770354071e-06, "loss": 0.0086, "step": 14362 }, { "epoch": 4.46, "learning_rate": 4.258451690338068e-06, "loss": 0.1069, "step": 14363 }, { "epoch": 4.46, "learning_rate": 4.258051610322065e-06, "loss": 0.0305, "step": 14364 }, { "epoch": 4.46, "learning_rate": 4.257651530306062e-06, "loss": 0.0159, "step": 14365 }, { "epoch": 4.46, "learning_rate": 4.2572514502900585e-06, "loss": 0.0085, "step": 14366 }, { "epoch": 4.46, "learning_rate": 4.256851370274055e-06, "loss": 0.0076, "step": 14367 }, { "epoch": 4.46, "learning_rate": 4.256451290258052e-06, "loss": 0.0112, "step": 14368 }, { "epoch": 4.46, "learning_rate": 4.256051210242049e-06, "loss": 0.0087, "step": 14369 }, { "epoch": 4.46, "learning_rate": 4.255651130226045e-06, "loss": 0.0263, "step": 14370 }, { "epoch": 4.46, "learning_rate": 4.255251050210043e-06, "loss": 0.0174, "step": 14371 }, { "epoch": 4.46, "learning_rate": 4.2548509701940395e-06, "loss": 0.0175, "step": 14372 }, { "epoch": 4.46, "learning_rate": 4.254450890178035e-06, "loss": 0.0228, "step": 14373 }, { "epoch": 4.46, "learning_rate": 4.254050810162033e-06, "loss": 0.0072, "step": 14374 }, { "epoch": 4.46, "learning_rate": 4.2536507301460296e-06, "loss": 0.0034, "step": 14375 }, { "epoch": 4.46, "learning_rate": 4.253250650130026e-06, "loss": 0.0072, "step": 14376 }, { "epoch": 4.46, "learning_rate": 4.252850570114023e-06, "loss": 0.0046, "step": 14377 }, { "epoch": 4.46, "learning_rate": 4.2524504900980205e-06, "loss": 0.0144, "step": 14378 }, { "epoch": 4.46, "learning_rate": 4.252050410082017e-06, "loss": 0.017, "step": 14379 }, { "epoch": 4.46, "learning_rate": 4.251650330066013e-06, "loss": 0.0173, "step": 14380 }, { "epoch": 4.46, "learning_rate": 4.2512502500500105e-06, "loss": 0.0166, "step": 14381 }, { "epoch": 4.46, "learning_rate": 4.250850170034007e-06, "loss": 0.0184, "step": 14382 }, { "epoch": 4.46, "learning_rate": 4.250450090018004e-06, "loss": 0.0426, "step": 14383 }, { "epoch": 4.46, "learning_rate": 4.250050010002001e-06, "loss": 0.0991, "step": 14384 }, { "epoch": 4.46, "learning_rate": 4.249649929985998e-06, "loss": 0.0127, "step": 14385 }, { "epoch": 4.47, "learning_rate": 4.249249849969994e-06, "loss": 0.1101, "step": 14386 }, { "epoch": 4.47, "learning_rate": 4.248849769953991e-06, "loss": 0.0183, "step": 14387 }, { "epoch": 4.47, "learning_rate": 4.248449689937988e-06, "loss": 0.0348, "step": 14388 }, { "epoch": 4.47, "learning_rate": 4.248049609921985e-06, "loss": 0.0208, "step": 14389 }, { "epoch": 4.47, "learning_rate": 4.2476495299059816e-06, "loss": 0.0079, "step": 14390 }, { "epoch": 4.47, "learning_rate": 4.247249449889978e-06, "loss": 0.0119, "step": 14391 }, { "epoch": 4.47, "learning_rate": 4.246849369873975e-06, "loss": 0.0334, "step": 14392 }, { "epoch": 4.47, "learning_rate": 4.246449289857972e-06, "loss": 0.0098, "step": 14393 }, { "epoch": 4.47, "learning_rate": 4.246049209841968e-06, "loss": 0.0182, "step": 14394 }, { "epoch": 4.47, "learning_rate": 4.245649129825966e-06, "loss": 0.0256, "step": 14395 }, { "epoch": 4.47, "learning_rate": 4.2452490498099625e-06, "loss": 0.0136, "step": 14396 }, { "epoch": 4.47, "learning_rate": 4.244848969793959e-06, "loss": 0.0177, "step": 14397 }, { "epoch": 4.47, "learning_rate": 4.244448889777956e-06, "loss": 0.0201, "step": 14398 }, { "epoch": 4.47, "learning_rate": 4.244048809761953e-06, "loss": 0.0193, "step": 14399 }, { "epoch": 4.47, "learning_rate": 4.243648729745949e-06, "loss": 0.0557, "step": 14400 }, { "epoch": 4.47, "learning_rate": 4.243248649729946e-06, "loss": 0.0251, "step": 14401 }, { "epoch": 4.47, "learning_rate": 4.2428485697139435e-06, "loss": 0.0201, "step": 14402 }, { "epoch": 4.47, "learning_rate": 4.24244848969794e-06, "loss": 0.0123, "step": 14403 }, { "epoch": 4.47, "learning_rate": 4.242048409681937e-06, "loss": 0.0717, "step": 14404 }, { "epoch": 4.47, "learning_rate": 4.2416483296659335e-06, "loss": 0.0184, "step": 14405 }, { "epoch": 4.47, "learning_rate": 4.24124824964993e-06, "loss": 0.0061, "step": 14406 }, { "epoch": 4.47, "learning_rate": 4.240848169633927e-06, "loss": 0.0091, "step": 14407 }, { "epoch": 4.47, "learning_rate": 4.240448089617924e-06, "loss": 0.0317, "step": 14408 }, { "epoch": 4.47, "learning_rate": 4.240048009601921e-06, "loss": 0.0084, "step": 14409 }, { "epoch": 4.47, "learning_rate": 4.239647929585918e-06, "loss": 0.049, "step": 14410 }, { "epoch": 4.47, "learning_rate": 4.239247849569914e-06, "loss": 0.0169, "step": 14411 }, { "epoch": 4.47, "learning_rate": 4.238847769553911e-06, "loss": 0.0106, "step": 14412 }, { "epoch": 4.47, "learning_rate": 4.238447689537908e-06, "loss": 0.0193, "step": 14413 }, { "epoch": 4.47, "learning_rate": 4.2380476095219046e-06, "loss": 0.0249, "step": 14414 }, { "epoch": 4.47, "learning_rate": 4.237647529505901e-06, "loss": 0.0107, "step": 14415 }, { "epoch": 4.47, "learning_rate": 4.237247449489899e-06, "loss": 0.014, "step": 14416 }, { "epoch": 4.47, "learning_rate": 4.236847369473895e-06, "loss": 0.011, "step": 14417 }, { "epoch": 4.48, "learning_rate": 4.236447289457892e-06, "loss": 0.0086, "step": 14418 }, { "epoch": 4.48, "learning_rate": 4.236047209441889e-06, "loss": 0.0038, "step": 14419 }, { "epoch": 4.48, "learning_rate": 4.2356471294258855e-06, "loss": 0.0251, "step": 14420 }, { "epoch": 4.48, "learning_rate": 4.235247049409882e-06, "loss": 0.0231, "step": 14421 }, { "epoch": 4.48, "learning_rate": 4.234846969393879e-06, "loss": 0.0693, "step": 14422 }, { "epoch": 4.48, "learning_rate": 4.2344468893778764e-06, "loss": 0.0347, "step": 14423 }, { "epoch": 4.48, "learning_rate": 4.234046809361872e-06, "loss": 0.0157, "step": 14424 }, { "epoch": 4.48, "learning_rate": 4.23364672934587e-06, "loss": 0.108, "step": 14425 }, { "epoch": 4.48, "learning_rate": 4.2332466493298665e-06, "loss": 0.0093, "step": 14426 }, { "epoch": 4.48, "learning_rate": 4.232846569313863e-06, "loss": 0.0158, "step": 14427 }, { "epoch": 4.48, "learning_rate": 4.23244648929786e-06, "loss": 0.0272, "step": 14428 }, { "epoch": 4.48, "learning_rate": 4.2320464092818566e-06, "loss": 0.0121, "step": 14429 }, { "epoch": 4.48, "learning_rate": 4.231646329265853e-06, "loss": 0.0349, "step": 14430 }, { "epoch": 4.48, "learning_rate": 4.23124624924985e-06, "loss": 0.0165, "step": 14431 }, { "epoch": 4.48, "learning_rate": 4.2308461692338475e-06, "loss": 0.0287, "step": 14432 }, { "epoch": 4.48, "learning_rate": 4.230446089217844e-06, "loss": 0.0382, "step": 14433 }, { "epoch": 4.48, "learning_rate": 4.230046009201841e-06, "loss": 0.0098, "step": 14434 }, { "epoch": 4.48, "learning_rate": 4.2296459291858375e-06, "loss": 0.0067, "step": 14435 }, { "epoch": 4.48, "learning_rate": 4.229245849169834e-06, "loss": 0.0145, "step": 14436 }, { "epoch": 4.48, "learning_rate": 4.228845769153831e-06, "loss": 0.0416, "step": 14437 }, { "epoch": 4.48, "learning_rate": 4.228445689137828e-06, "loss": 0.0251, "step": 14438 }, { "epoch": 4.48, "learning_rate": 4.228045609121825e-06, "loss": 0.0272, "step": 14439 }, { "epoch": 4.48, "learning_rate": 4.227645529105822e-06, "loss": 0.0069, "step": 14440 }, { "epoch": 4.48, "learning_rate": 4.2272454490898185e-06, "loss": 0.058, "step": 14441 }, { "epoch": 4.48, "learning_rate": 4.226845369073815e-06, "loss": 0.0228, "step": 14442 }, { "epoch": 4.48, "learning_rate": 4.226445289057812e-06, "loss": 0.0058, "step": 14443 }, { "epoch": 4.48, "learning_rate": 4.2260452090418085e-06, "loss": 0.0447, "step": 14444 }, { "epoch": 4.48, "learning_rate": 4.225645129025805e-06, "loss": 0.0092, "step": 14445 }, { "epoch": 4.48, "learning_rate": 4.225245049009803e-06, "loss": 0.1403, "step": 14446 }, { "epoch": 4.48, "learning_rate": 4.2248449689937995e-06, "loss": 0.009, "step": 14447 }, { "epoch": 4.48, "learning_rate": 4.224444888977796e-06, "loss": 0.0174, "step": 14448 }, { "epoch": 4.48, "learning_rate": 4.224044808961793e-06, "loss": 0.0091, "step": 14449 }, { "epoch": 4.48, "learning_rate": 4.2236447289457895e-06, "loss": 0.0192, "step": 14450 }, { "epoch": 4.49, "learning_rate": 4.223244648929786e-06, "loss": 0.0211, "step": 14451 }, { "epoch": 4.49, "learning_rate": 4.222844568913783e-06, "loss": 0.0186, "step": 14452 }, { "epoch": 4.49, "learning_rate": 4.22244448889778e-06, "loss": 0.0078, "step": 14453 }, { "epoch": 4.49, "learning_rate": 4.222044408881777e-06, "loss": 0.0088, "step": 14454 }, { "epoch": 4.49, "learning_rate": 4.221644328865773e-06, "loss": 0.04, "step": 14455 }, { "epoch": 4.49, "learning_rate": 4.2212442488497705e-06, "loss": 0.0431, "step": 14456 }, { "epoch": 4.49, "learning_rate": 4.220844168833767e-06, "loss": 0.0219, "step": 14457 }, { "epoch": 4.49, "learning_rate": 4.220444088817764e-06, "loss": 0.0661, "step": 14458 }, { "epoch": 4.49, "learning_rate": 4.2200440088017605e-06, "loss": 0.0314, "step": 14459 }, { "epoch": 4.49, "learning_rate": 4.219643928785758e-06, "loss": 0.0159, "step": 14460 }, { "epoch": 4.49, "learning_rate": 4.219243848769755e-06, "loss": 0.0269, "step": 14461 }, { "epoch": 4.49, "learning_rate": 4.218843768753751e-06, "loss": 0.0266, "step": 14462 }, { "epoch": 4.49, "learning_rate": 4.218443688737748e-06, "loss": 0.0497, "step": 14463 }, { "epoch": 4.49, "learning_rate": 4.218043608721745e-06, "loss": 0.0185, "step": 14464 }, { "epoch": 4.49, "learning_rate": 4.2176435287057415e-06, "loss": 0.0062, "step": 14465 }, { "epoch": 4.49, "learning_rate": 4.217243448689738e-06, "loss": 0.039, "step": 14466 }, { "epoch": 4.49, "learning_rate": 4.216843368673736e-06, "loss": 0.0114, "step": 14467 }, { "epoch": 4.49, "learning_rate": 4.2164432886577316e-06, "loss": 0.0106, "step": 14468 }, { "epoch": 4.49, "learning_rate": 4.216043208641728e-06, "loss": 0.0451, "step": 14469 }, { "epoch": 4.49, "learning_rate": 4.215643128625726e-06, "loss": 0.0434, "step": 14470 }, { "epoch": 4.49, "learning_rate": 4.2152430486097225e-06, "loss": 0.0154, "step": 14471 }, { "epoch": 4.49, "learning_rate": 4.214842968593719e-06, "loss": 0.0135, "step": 14472 }, { "epoch": 4.49, "learning_rate": 4.214442888577716e-06, "loss": 0.0284, "step": 14473 }, { "epoch": 4.49, "learning_rate": 4.2140428085617125e-06, "loss": 0.0348, "step": 14474 }, { "epoch": 4.49, "learning_rate": 4.213642728545709e-06, "loss": 0.0124, "step": 14475 }, { "epoch": 4.49, "learning_rate": 4.213242648529706e-06, "loss": 0.0169, "step": 14476 }, { "epoch": 4.49, "learning_rate": 4.2128425685137034e-06, "loss": 0.0222, "step": 14477 }, { "epoch": 4.49, "learning_rate": 4.2124424884977e-06, "loss": 0.0502, "step": 14478 }, { "epoch": 4.49, "learning_rate": 4.212042408481697e-06, "loss": 0.077, "step": 14479 }, { "epoch": 4.49, "learning_rate": 4.2116423284656935e-06, "loss": 0.0212, "step": 14480 }, { "epoch": 4.49, "learning_rate": 4.21124224844969e-06, "loss": 0.0671, "step": 14481 }, { "epoch": 4.49, "learning_rate": 4.210842168433687e-06, "loss": 0.0058, "step": 14482 }, { "epoch": 4.5, "learning_rate": 4.2104420884176835e-06, "loss": 0.0314, "step": 14483 }, { "epoch": 4.5, "learning_rate": 4.210042008401681e-06, "loss": 0.0098, "step": 14484 }, { "epoch": 4.5, "learning_rate": 4.209641928385678e-06, "loss": 0.03, "step": 14485 }, { "epoch": 4.5, "learning_rate": 4.2092418483696745e-06, "loss": 0.0458, "step": 14486 }, { "epoch": 4.5, "learning_rate": 4.208841768353671e-06, "loss": 0.015, "step": 14487 }, { "epoch": 4.5, "learning_rate": 4.208441688337668e-06, "loss": 0.0075, "step": 14488 }, { "epoch": 4.5, "learning_rate": 4.2080416083216645e-06, "loss": 0.036, "step": 14489 }, { "epoch": 4.5, "learning_rate": 4.207641528305661e-06, "loss": 0.0286, "step": 14490 }, { "epoch": 4.5, "learning_rate": 4.207241448289659e-06, "loss": 0.0228, "step": 14491 }, { "epoch": 4.5, "learning_rate": 4.206841368273655e-06, "loss": 0.014, "step": 14492 }, { "epoch": 4.5, "learning_rate": 4.206441288257651e-06, "loss": 0.0285, "step": 14493 }, { "epoch": 4.5, "learning_rate": 4.206041208241649e-06, "loss": 0.0784, "step": 14494 }, { "epoch": 4.5, "learning_rate": 4.2056411282256455e-06, "loss": 0.0181, "step": 14495 }, { "epoch": 4.5, "learning_rate": 4.205241048209642e-06, "loss": 0.0068, "step": 14496 }, { "epoch": 4.5, "learning_rate": 4.204840968193639e-06, "loss": 0.0329, "step": 14497 }, { "epoch": 4.5, "learning_rate": 4.204440888177636e-06, "loss": 0.0312, "step": 14498 }, { "epoch": 4.5, "learning_rate": 4.204040808161632e-06, "loss": 0.0312, "step": 14499 }, { "epoch": 4.5, "learning_rate": 4.203640728145629e-06, "loss": 0.0247, "step": 14500 }, { "epoch": 4.5, "learning_rate": 4.2032406481296264e-06, "loss": 0.1038, "step": 14501 }, { "epoch": 4.5, "learning_rate": 4.202840568113623e-06, "loss": 0.0471, "step": 14502 }, { "epoch": 4.5, "learning_rate": 4.20244048809762e-06, "loss": 0.0179, "step": 14503 }, { "epoch": 4.5, "learning_rate": 4.2020404080816165e-06, "loss": 0.022, "step": 14504 }, { "epoch": 4.5, "learning_rate": 4.201640328065614e-06, "loss": 0.0375, "step": 14505 }, { "epoch": 4.5, "learning_rate": 4.20124024804961e-06, "loss": 0.0145, "step": 14506 }, { "epoch": 4.5, "learning_rate": 4.2008401680336066e-06, "loss": 0.0214, "step": 14507 }, { "epoch": 4.5, "learning_rate": 4.200440088017604e-06, "loss": 0.0974, "step": 14508 }, { "epoch": 4.5, "learning_rate": 4.200040008001601e-06, "loss": 0.0989, "step": 14509 }, { "epoch": 4.5, "learning_rate": 4.1996399279855975e-06, "loss": 0.0127, "step": 14510 }, { "epoch": 4.5, "learning_rate": 4.199239847969594e-06, "loss": 0.0169, "step": 14511 }, { "epoch": 4.5, "learning_rate": 4.198839767953591e-06, "loss": 0.0253, "step": 14512 }, { "epoch": 4.5, "learning_rate": 4.1984396879375875e-06, "loss": 0.0154, "step": 14513 }, { "epoch": 4.5, "learning_rate": 4.198039607921584e-06, "loss": 0.0099, "step": 14514 }, { "epoch": 4.51, "learning_rate": 4.197639527905582e-06, "loss": 0.0056, "step": 14515 }, { "epoch": 4.51, "learning_rate": 4.1972394478895784e-06, "loss": 0.0322, "step": 14516 }, { "epoch": 4.51, "learning_rate": 4.196839367873575e-06, "loss": 0.0612, "step": 14517 }, { "epoch": 4.51, "learning_rate": 4.196439287857572e-06, "loss": 0.0463, "step": 14518 }, { "epoch": 4.51, "learning_rate": 4.1960392078415685e-06, "loss": 0.1071, "step": 14519 }, { "epoch": 4.51, "learning_rate": 4.195639127825565e-06, "loss": 0.0407, "step": 14520 }, { "epoch": 4.51, "learning_rate": 4.195239047809562e-06, "loss": 0.0095, "step": 14521 }, { "epoch": 4.51, "learning_rate": 4.194838967793559e-06, "loss": 0.0118, "step": 14522 }, { "epoch": 4.51, "learning_rate": 4.194438887777556e-06, "loss": 0.0741, "step": 14523 }, { "epoch": 4.51, "learning_rate": 4.194038807761553e-06, "loss": 0.0074, "step": 14524 }, { "epoch": 4.51, "learning_rate": 4.1936387277455495e-06, "loss": 0.0248, "step": 14525 }, { "epoch": 4.51, "learning_rate": 4.193238647729546e-06, "loss": 0.1242, "step": 14526 }, { "epoch": 4.51, "learning_rate": 4.192838567713543e-06, "loss": 0.0215, "step": 14527 }, { "epoch": 4.51, "learning_rate": 4.1924384876975395e-06, "loss": 0.0184, "step": 14528 }, { "epoch": 4.51, "learning_rate": 4.192038407681537e-06, "loss": 0.0812, "step": 14529 }, { "epoch": 4.51, "learning_rate": 4.191638327665534e-06, "loss": 0.0253, "step": 14530 }, { "epoch": 4.51, "learning_rate": 4.19123824764953e-06, "loss": 0.0065, "step": 14531 }, { "epoch": 4.51, "learning_rate": 4.190838167633527e-06, "loss": 0.0687, "step": 14532 }, { "epoch": 4.51, "learning_rate": 4.190438087617524e-06, "loss": 0.0062, "step": 14533 }, { "epoch": 4.51, "learning_rate": 4.1900380076015205e-06, "loss": 0.0103, "step": 14534 }, { "epoch": 4.51, "learning_rate": 4.189637927585517e-06, "loss": 0.0122, "step": 14535 }, { "epoch": 4.51, "learning_rate": 4.189237847569515e-06, "loss": 0.0147, "step": 14536 }, { "epoch": 4.51, "learning_rate": 4.1888377675535105e-06, "loss": 0.0218, "step": 14537 }, { "epoch": 4.51, "learning_rate": 4.188437687537508e-06, "loss": 0.009, "step": 14538 }, { "epoch": 4.51, "learning_rate": 4.188037607521505e-06, "loss": 0.0285, "step": 14539 }, { "epoch": 4.51, "learning_rate": 4.1876375275055014e-06, "loss": 0.0305, "step": 14540 }, { "epoch": 4.51, "learning_rate": 4.187237447489498e-06, "loss": 0.0054, "step": 14541 }, { "epoch": 4.51, "learning_rate": 4.186837367473495e-06, "loss": 0.0781, "step": 14542 }, { "epoch": 4.51, "learning_rate": 4.186437287457492e-06, "loss": 0.0168, "step": 14543 }, { "epoch": 4.51, "learning_rate": 4.186037207441488e-06, "loss": 0.0129, "step": 14544 }, { "epoch": 4.51, "learning_rate": 4.185637127425486e-06, "loss": 0.0089, "step": 14545 }, { "epoch": 4.51, "learning_rate": 4.185237047409482e-06, "loss": 0.0166, "step": 14546 }, { "epoch": 4.52, "learning_rate": 4.184836967393479e-06, "loss": 0.0075, "step": 14547 }, { "epoch": 4.52, "learning_rate": 4.184436887377476e-06, "loss": 0.0113, "step": 14548 }, { "epoch": 4.52, "learning_rate": 4.1840368073614725e-06, "loss": 0.0205, "step": 14549 }, { "epoch": 4.52, "learning_rate": 4.183636727345469e-06, "loss": 0.042, "step": 14550 }, { "epoch": 4.52, "learning_rate": 4.183236647329466e-06, "loss": 0.0264, "step": 14551 }, { "epoch": 4.52, "learning_rate": 4.182836567313463e-06, "loss": 0.0107, "step": 14552 }, { "epoch": 4.52, "learning_rate": 4.18243648729746e-06, "loss": 0.039, "step": 14553 }, { "epoch": 4.52, "learning_rate": 4.182036407281457e-06, "loss": 0.0079, "step": 14554 }, { "epoch": 4.52, "learning_rate": 4.1816363272654534e-06, "loss": 0.0159, "step": 14555 }, { "epoch": 4.52, "learning_rate": 4.18123624724945e-06, "loss": 0.0202, "step": 14556 }, { "epoch": 4.52, "learning_rate": 4.180836167233447e-06, "loss": 0.0039, "step": 14557 }, { "epoch": 4.52, "learning_rate": 4.1804360872174435e-06, "loss": 0.017, "step": 14558 }, { "epoch": 4.52, "learning_rate": 4.180036007201441e-06, "loss": 0.0182, "step": 14559 }, { "epoch": 4.52, "learning_rate": 4.179635927185438e-06, "loss": 0.0068, "step": 14560 }, { "epoch": 4.52, "learning_rate": 4.179235847169434e-06, "loss": 0.0121, "step": 14561 }, { "epoch": 4.52, "learning_rate": 4.178835767153431e-06, "loss": 0.0073, "step": 14562 }, { "epoch": 4.52, "learning_rate": 4.178435687137428e-06, "loss": 0.0408, "step": 14563 }, { "epoch": 4.52, "learning_rate": 4.1780356071214245e-06, "loss": 0.0077, "step": 14564 }, { "epoch": 4.52, "learning_rate": 4.177635527105421e-06, "loss": 0.0322, "step": 14565 }, { "epoch": 4.52, "learning_rate": 4.177235447089419e-06, "loss": 0.015, "step": 14566 }, { "epoch": 4.52, "learning_rate": 4.176835367073415e-06, "loss": 0.023, "step": 14567 }, { "epoch": 4.52, "learning_rate": 4.176435287057412e-06, "loss": 0.0187, "step": 14568 }, { "epoch": 4.52, "learning_rate": 4.176035207041409e-06, "loss": 0.0104, "step": 14569 }, { "epoch": 4.52, "learning_rate": 4.175635127025405e-06, "loss": 0.0154, "step": 14570 }, { "epoch": 4.52, "learning_rate": 4.175235047009402e-06, "loss": 0.0165, "step": 14571 }, { "epoch": 4.52, "learning_rate": 4.174834966993399e-06, "loss": 0.0135, "step": 14572 }, { "epoch": 4.52, "learning_rate": 4.174434886977396e-06, "loss": 0.0241, "step": 14573 }, { "epoch": 4.52, "learning_rate": 4.174034806961393e-06, "loss": 0.0067, "step": 14574 }, { "epoch": 4.52, "learning_rate": 4.173634726945389e-06, "loss": 0.0139, "step": 14575 }, { "epoch": 4.52, "learning_rate": 4.173234646929386e-06, "loss": 0.0181, "step": 14576 }, { "epoch": 4.52, "learning_rate": 4.172834566913383e-06, "loss": 0.0216, "step": 14577 }, { "epoch": 4.52, "learning_rate": 4.17243448689738e-06, "loss": 0.0327, "step": 14578 }, { "epoch": 4.53, "learning_rate": 4.1720344068813764e-06, "loss": 0.0275, "step": 14579 }, { "epoch": 4.53, "learning_rate": 4.171634326865374e-06, "loss": 0.0109, "step": 14580 }, { "epoch": 4.53, "learning_rate": 4.17123424684937e-06, "loss": 0.024, "step": 14581 }, { "epoch": 4.53, "learning_rate": 4.1708341668333665e-06, "loss": 0.0167, "step": 14582 }, { "epoch": 4.53, "learning_rate": 4.170434086817364e-06, "loss": 0.0153, "step": 14583 }, { "epoch": 4.53, "learning_rate": 4.170034006801361e-06, "loss": 0.0213, "step": 14584 }, { "epoch": 4.53, "learning_rate": 4.169633926785357e-06, "loss": 0.0867, "step": 14585 }, { "epoch": 4.53, "learning_rate": 4.169233846769354e-06, "loss": 0.0156, "step": 14586 }, { "epoch": 4.53, "learning_rate": 4.168833766753352e-06, "loss": 0.0139, "step": 14587 }, { "epoch": 4.53, "learning_rate": 4.1684336867373475e-06, "loss": 0.0159, "step": 14588 }, { "epoch": 4.53, "learning_rate": 4.168033606721344e-06, "loss": 0.0184, "step": 14589 }, { "epoch": 4.53, "learning_rate": 4.167633526705342e-06, "loss": 0.1074, "step": 14590 }, { "epoch": 4.53, "learning_rate": 4.167233446689338e-06, "loss": 0.0121, "step": 14591 }, { "epoch": 4.53, "learning_rate": 4.166833366673335e-06, "loss": 0.0146, "step": 14592 }, { "epoch": 4.53, "learning_rate": 4.166433286657332e-06, "loss": 0.0278, "step": 14593 }, { "epoch": 4.53, "learning_rate": 4.1660332066413284e-06, "loss": 0.0139, "step": 14594 }, { "epoch": 4.53, "learning_rate": 4.165633126625325e-06, "loss": 0.0114, "step": 14595 }, { "epoch": 4.53, "learning_rate": 4.165233046609322e-06, "loss": 0.0145, "step": 14596 }, { "epoch": 4.53, "learning_rate": 4.164832966593319e-06, "loss": 0.0197, "step": 14597 }, { "epoch": 4.53, "learning_rate": 4.164432886577316e-06, "loss": 0.0288, "step": 14598 }, { "epoch": 4.53, "learning_rate": 4.164032806561313e-06, "loss": 0.0796, "step": 14599 }, { "epoch": 4.53, "learning_rate": 4.163632726545309e-06, "loss": 0.0142, "step": 14600 }, { "epoch": 4.53, "learning_rate": 4.163232646529306e-06, "loss": 0.0358, "step": 14601 }, { "epoch": 4.53, "learning_rate": 4.162832566513303e-06, "loss": 0.0224, "step": 14602 }, { "epoch": 4.53, "learning_rate": 4.1624324864972995e-06, "loss": 0.0164, "step": 14603 }, { "epoch": 4.53, "learning_rate": 4.162032406481297e-06, "loss": 0.0115, "step": 14604 }, { "epoch": 4.53, "learning_rate": 4.161632326465294e-06, "loss": 0.0672, "step": 14605 }, { "epoch": 4.53, "learning_rate": 4.16123224644929e-06, "loss": 0.0262, "step": 14606 }, { "epoch": 4.53, "learning_rate": 4.160832166433287e-06, "loss": 0.0112, "step": 14607 }, { "epoch": 4.53, "learning_rate": 4.160432086417284e-06, "loss": 0.0191, "step": 14608 }, { "epoch": 4.53, "learning_rate": 4.16003200640128e-06, "loss": 0.0231, "step": 14609 }, { "epoch": 4.53, "learning_rate": 4.159631926385277e-06, "loss": 0.0165, "step": 14610 }, { "epoch": 4.53, "learning_rate": 4.159231846369275e-06, "loss": 0.023, "step": 14611 }, { "epoch": 4.54, "learning_rate": 4.158831766353271e-06, "loss": 0.0112, "step": 14612 }, { "epoch": 4.54, "learning_rate": 4.158431686337267e-06, "loss": 0.007, "step": 14613 }, { "epoch": 4.54, "learning_rate": 4.158031606321265e-06, "loss": 0.036, "step": 14614 }, { "epoch": 4.54, "learning_rate": 4.157631526305261e-06, "loss": 0.0173, "step": 14615 }, { "epoch": 4.54, "learning_rate": 4.157231446289258e-06, "loss": 0.0582, "step": 14616 }, { "epoch": 4.54, "learning_rate": 4.156831366273255e-06, "loss": 0.0553, "step": 14617 }, { "epoch": 4.54, "learning_rate": 4.156431286257252e-06, "loss": 0.024, "step": 14618 }, { "epoch": 4.54, "learning_rate": 4.156031206241248e-06, "loss": 0.0222, "step": 14619 }, { "epoch": 4.54, "learning_rate": 4.155631126225245e-06, "loss": 0.0918, "step": 14620 }, { "epoch": 4.54, "learning_rate": 4.155231046209242e-06, "loss": 0.0156, "step": 14621 }, { "epoch": 4.54, "learning_rate": 4.154830966193239e-06, "loss": 0.0313, "step": 14622 }, { "epoch": 4.54, "learning_rate": 4.154430886177236e-06, "loss": 0.0992, "step": 14623 }, { "epoch": 4.54, "learning_rate": 4.154030806161232e-06, "loss": 0.0292, "step": 14624 }, { "epoch": 4.54, "learning_rate": 4.153630726145229e-06, "loss": 0.0092, "step": 14625 }, { "epoch": 4.54, "learning_rate": 4.153230646129226e-06, "loss": 0.074, "step": 14626 }, { "epoch": 4.54, "learning_rate": 4.1528305661132225e-06, "loss": 0.006, "step": 14627 }, { "epoch": 4.54, "learning_rate": 4.15243048609722e-06, "loss": 0.0048, "step": 14628 }, { "epoch": 4.54, "learning_rate": 4.152030406081217e-06, "loss": 0.0093, "step": 14629 }, { "epoch": 4.54, "learning_rate": 4.151630326065213e-06, "loss": 0.0104, "step": 14630 }, { "epoch": 4.54, "learning_rate": 4.15123024604921e-06, "loss": 0.0036, "step": 14631 }, { "epoch": 4.54, "learning_rate": 4.150830166033207e-06, "loss": 0.0166, "step": 14632 }, { "epoch": 4.54, "learning_rate": 4.1504300860172034e-06, "loss": 0.0213, "step": 14633 }, { "epoch": 4.54, "learning_rate": 4.150030006001201e-06, "loss": 0.0215, "step": 14634 }, { "epoch": 4.54, "learning_rate": 4.149629925985198e-06, "loss": 0.0723, "step": 14635 }, { "epoch": 4.54, "learning_rate": 4.149229845969194e-06, "loss": 0.0473, "step": 14636 }, { "epoch": 4.54, "learning_rate": 4.148829765953191e-06, "loss": 0.0133, "step": 14637 }, { "epoch": 4.54, "learning_rate": 4.148429685937188e-06, "loss": 0.0061, "step": 14638 }, { "epoch": 4.54, "learning_rate": 4.148029605921184e-06, "loss": 0.0089, "step": 14639 }, { "epoch": 4.54, "learning_rate": 4.147629525905181e-06, "loss": 0.0163, "step": 14640 }, { "epoch": 4.54, "learning_rate": 4.147229445889179e-06, "loss": 0.0212, "step": 14641 }, { "epoch": 4.54, "learning_rate": 4.146829365873175e-06, "loss": 0.0483, "step": 14642 }, { "epoch": 4.54, "learning_rate": 4.146429285857172e-06, "loss": 0.0072, "step": 14643 }, { "epoch": 4.55, "learning_rate": 4.146029205841169e-06, "loss": 0.0069, "step": 14644 }, { "epoch": 4.55, "learning_rate": 4.145629125825165e-06, "loss": 0.0259, "step": 14645 }, { "epoch": 4.55, "learning_rate": 4.145229045809162e-06, "loss": 0.0106, "step": 14646 }, { "epoch": 4.55, "learning_rate": 4.144828965793159e-06, "loss": 0.0076, "step": 14647 }, { "epoch": 4.55, "learning_rate": 4.144428885777156e-06, "loss": 0.0817, "step": 14648 }, { "epoch": 4.55, "learning_rate": 4.144028805761153e-06, "loss": 0.0186, "step": 14649 }, { "epoch": 4.55, "learning_rate": 4.14362872574515e-06, "loss": 0.0289, "step": 14650 }, { "epoch": 4.55, "learning_rate": 4.143228645729146e-06, "loss": 0.0364, "step": 14651 }, { "epoch": 4.55, "learning_rate": 4.142828565713143e-06, "loss": 0.0154, "step": 14652 }, { "epoch": 4.55, "learning_rate": 4.14242848569714e-06, "loss": 0.0077, "step": 14653 }, { "epoch": 4.55, "learning_rate": 4.142028405681136e-06, "loss": 0.0143, "step": 14654 }, { "epoch": 4.55, "learning_rate": 4.141628325665134e-06, "loss": 0.0152, "step": 14655 }, { "epoch": 4.55, "learning_rate": 4.141228245649131e-06, "loss": 0.0458, "step": 14656 }, { "epoch": 4.55, "learning_rate": 4.1408281656331264e-06, "loss": 0.02, "step": 14657 }, { "epoch": 4.55, "learning_rate": 4.140428085617124e-06, "loss": 0.0134, "step": 14658 }, { "epoch": 4.55, "learning_rate": 4.140028005601121e-06, "loss": 0.0787, "step": 14659 }, { "epoch": 4.55, "learning_rate": 4.139627925585117e-06, "loss": 0.0234, "step": 14660 }, { "epoch": 4.55, "learning_rate": 4.139227845569114e-06, "loss": 0.013, "step": 14661 }, { "epoch": 4.55, "learning_rate": 4.1388277655531116e-06, "loss": 0.0207, "step": 14662 }, { "epoch": 4.55, "learning_rate": 4.138427685537107e-06, "loss": 0.0068, "step": 14663 }, { "epoch": 4.55, "learning_rate": 4.138027605521104e-06, "loss": 0.0093, "step": 14664 }, { "epoch": 4.55, "learning_rate": 4.137627525505102e-06, "loss": 0.0107, "step": 14665 }, { "epoch": 4.55, "learning_rate": 4.137227445489098e-06, "loss": 0.026, "step": 14666 }, { "epoch": 4.55, "learning_rate": 4.136827365473095e-06, "loss": 0.0139, "step": 14667 }, { "epoch": 4.55, "learning_rate": 4.136427285457092e-06, "loss": 0.0161, "step": 14668 }, { "epoch": 4.55, "learning_rate": 4.136027205441089e-06, "loss": 0.0888, "step": 14669 }, { "epoch": 4.55, "learning_rate": 4.135627125425085e-06, "loss": 0.0149, "step": 14670 }, { "epoch": 4.55, "learning_rate": 4.135227045409082e-06, "loss": 0.0407, "step": 14671 }, { "epoch": 4.55, "learning_rate": 4.134826965393079e-06, "loss": 0.0301, "step": 14672 }, { "epoch": 4.55, "learning_rate": 4.134426885377076e-06, "loss": 0.0236, "step": 14673 }, { "epoch": 4.55, "learning_rate": 4.134026805361073e-06, "loss": 0.0167, "step": 14674 }, { "epoch": 4.55, "learning_rate": 4.133626725345069e-06, "loss": 0.0474, "step": 14675 }, { "epoch": 4.56, "learning_rate": 4.133226645329066e-06, "loss": 0.006, "step": 14676 }, { "epoch": 4.56, "learning_rate": 4.132826565313063e-06, "loss": 0.0107, "step": 14677 }, { "epoch": 4.56, "learning_rate": 4.132426485297059e-06, "loss": 0.0089, "step": 14678 }, { "epoch": 4.56, "learning_rate": 4.132026405281057e-06, "loss": 0.0166, "step": 14679 }, { "epoch": 4.56, "learning_rate": 4.131626325265054e-06, "loss": 0.0205, "step": 14680 }, { "epoch": 4.56, "learning_rate": 4.13122624524905e-06, "loss": 0.0163, "step": 14681 }, { "epoch": 4.56, "learning_rate": 4.130826165233047e-06, "loss": 0.0078, "step": 14682 }, { "epoch": 4.56, "learning_rate": 4.130426085217044e-06, "loss": 0.0202, "step": 14683 }, { "epoch": 4.56, "learning_rate": 4.13002600520104e-06, "loss": 0.01, "step": 14684 }, { "epoch": 4.56, "learning_rate": 4.129625925185037e-06, "loss": 0.0109, "step": 14685 }, { "epoch": 4.56, "learning_rate": 4.1292258451690346e-06, "loss": 0.0072, "step": 14686 }, { "epoch": 4.56, "learning_rate": 4.128825765153031e-06, "loss": 0.0103, "step": 14687 }, { "epoch": 4.56, "learning_rate": 4.128425685137028e-06, "loss": 0.0189, "step": 14688 }, { "epoch": 4.56, "learning_rate": 4.128025605121025e-06, "loss": 0.0432, "step": 14689 }, { "epoch": 4.56, "learning_rate": 4.127625525105021e-06, "loss": 0.0153, "step": 14690 }, { "epoch": 4.56, "learning_rate": 4.127225445089018e-06, "loss": 0.0043, "step": 14691 }, { "epoch": 4.56, "learning_rate": 4.126825365073015e-06, "loss": 0.0205, "step": 14692 }, { "epoch": 4.56, "learning_rate": 4.126425285057012e-06, "loss": 0.0284, "step": 14693 }, { "epoch": 4.56, "learning_rate": 4.126025205041009e-06, "loss": 0.0203, "step": 14694 }, { "epoch": 4.56, "learning_rate": 4.125625125025005e-06, "loss": 0.0483, "step": 14695 }, { "epoch": 4.56, "learning_rate": 4.125225045009002e-06, "loss": 0.0218, "step": 14696 }, { "epoch": 4.56, "learning_rate": 4.124824964992999e-06, "loss": 0.0144, "step": 14697 }, { "epoch": 4.56, "learning_rate": 4.124424884976996e-06, "loss": 0.0711, "step": 14698 }, { "epoch": 4.56, "learning_rate": 4.124024804960992e-06, "loss": 0.0264, "step": 14699 }, { "epoch": 4.56, "learning_rate": 4.12362472494499e-06, "loss": 0.017, "step": 14700 }, { "epoch": 4.56, "learning_rate": 4.123224644928986e-06, "loss": 0.006, "step": 14701 }, { "epoch": 4.56, "learning_rate": 4.122824564912982e-06, "loss": 0.0086, "step": 14702 }, { "epoch": 4.56, "learning_rate": 4.12242448489698e-06, "loss": 0.0151, "step": 14703 }, { "epoch": 4.56, "learning_rate": 4.122024404880977e-06, "loss": 0.0272, "step": 14704 }, { "epoch": 4.56, "learning_rate": 4.121624324864973e-06, "loss": 0.0155, "step": 14705 }, { "epoch": 4.56, "learning_rate": 4.12122424484897e-06, "loss": 0.0139, "step": 14706 }, { "epoch": 4.56, "learning_rate": 4.120824164832967e-06, "loss": 0.0203, "step": 14707 }, { "epoch": 4.57, "learning_rate": 4.120424084816963e-06, "loss": 0.027, "step": 14708 }, { "epoch": 4.57, "learning_rate": 4.12002400480096e-06, "loss": 0.0173, "step": 14709 }, { "epoch": 4.57, "learning_rate": 4.119623924784958e-06, "loss": 0.0457, "step": 14710 }, { "epoch": 4.57, "learning_rate": 4.119223844768954e-06, "loss": 0.0107, "step": 14711 }, { "epoch": 4.57, "learning_rate": 4.118823764752951e-06, "loss": 0.013, "step": 14712 }, { "epoch": 4.57, "learning_rate": 4.118423684736948e-06, "loss": 0.0046, "step": 14713 }, { "epoch": 4.57, "learning_rate": 4.118023604720944e-06, "loss": 0.0056, "step": 14714 }, { "epoch": 4.57, "learning_rate": 4.117623524704941e-06, "loss": 0.1185, "step": 14715 }, { "epoch": 4.57, "learning_rate": 4.117223444688938e-06, "loss": 0.0119, "step": 14716 }, { "epoch": 4.57, "learning_rate": 4.116823364672935e-06, "loss": 0.0089, "step": 14717 }, { "epoch": 4.57, "learning_rate": 4.116423284656932e-06, "loss": 0.016, "step": 14718 }, { "epoch": 4.57, "learning_rate": 4.116023204640929e-06, "loss": 0.0067, "step": 14719 }, { "epoch": 4.57, "learning_rate": 4.115623124624925e-06, "loss": 0.0059, "step": 14720 }, { "epoch": 4.57, "learning_rate": 4.115223044608922e-06, "loss": 0.0243, "step": 14721 }, { "epoch": 4.57, "learning_rate": 4.114822964592919e-06, "loss": 0.0939, "step": 14722 }, { "epoch": 4.57, "learning_rate": 4.114422884576915e-06, "loss": 0.0298, "step": 14723 }, { "epoch": 4.57, "learning_rate": 4.114022804560913e-06, "loss": 0.0185, "step": 14724 }, { "epoch": 4.57, "learning_rate": 4.1136227245449096e-06, "loss": 0.0251, "step": 14725 }, { "epoch": 4.57, "learning_rate": 4.113222644528906e-06, "loss": 0.0145, "step": 14726 }, { "epoch": 4.57, "learning_rate": 4.112822564512903e-06, "loss": 0.018, "step": 14727 }, { "epoch": 4.57, "learning_rate": 4.1124224844969e-06, "loss": 0.0315, "step": 14728 }, { "epoch": 4.57, "learning_rate": 4.112022404480896e-06, "loss": 0.0348, "step": 14729 }, { "epoch": 4.57, "learning_rate": 4.111622324464893e-06, "loss": 0.0174, "step": 14730 }, { "epoch": 4.57, "learning_rate": 4.1112222444488905e-06, "loss": 0.041, "step": 14731 }, { "epoch": 4.57, "learning_rate": 4.110822164432887e-06, "loss": 0.0406, "step": 14732 }, { "epoch": 4.57, "learning_rate": 4.110422084416884e-06, "loss": 0.0163, "step": 14733 }, { "epoch": 4.57, "learning_rate": 4.110022004400881e-06, "loss": 0.1124, "step": 14734 }, { "epoch": 4.57, "learning_rate": 4.109621924384877e-06, "loss": 0.0309, "step": 14735 }, { "epoch": 4.57, "learning_rate": 4.109221844368874e-06, "loss": 0.0084, "step": 14736 }, { "epoch": 4.57, "learning_rate": 4.108821764352871e-06, "loss": 0.0042, "step": 14737 }, { "epoch": 4.57, "learning_rate": 4.108421684336868e-06, "loss": 0.0084, "step": 14738 }, { "epoch": 4.57, "learning_rate": 4.108021604320864e-06, "loss": 0.0107, "step": 14739 }, { "epoch": 4.57, "learning_rate": 4.1076215243048616e-06, "loss": 0.0171, "step": 14740 }, { "epoch": 4.58, "learning_rate": 4.107221444288858e-06, "loss": 0.0159, "step": 14741 }, { "epoch": 4.58, "learning_rate": 4.106821364272855e-06, "loss": 0.0712, "step": 14742 }, { "epoch": 4.58, "learning_rate": 4.106421284256852e-06, "loss": 0.0456, "step": 14743 }, { "epoch": 4.58, "learning_rate": 4.106021204240848e-06, "loss": 0.0229, "step": 14744 }, { "epoch": 4.58, "learning_rate": 4.105621124224845e-06, "loss": 0.0203, "step": 14745 }, { "epoch": 4.58, "learning_rate": 4.105221044208842e-06, "loss": 0.0084, "step": 14746 }, { "epoch": 4.58, "learning_rate": 4.104820964192839e-06, "loss": 0.0465, "step": 14747 }, { "epoch": 4.58, "learning_rate": 4.104420884176836e-06, "loss": 0.0078, "step": 14748 }, { "epoch": 4.58, "learning_rate": 4.104020804160833e-06, "loss": 0.0056, "step": 14749 }, { "epoch": 4.58, "learning_rate": 4.103620724144829e-06, "loss": 0.0185, "step": 14750 }, { "epoch": 4.58, "learning_rate": 4.103220644128826e-06, "loss": 0.0209, "step": 14751 }, { "epoch": 4.58, "learning_rate": 4.102820564112823e-06, "loss": 0.0111, "step": 14752 }, { "epoch": 4.58, "learning_rate": 4.102420484096819e-06, "loss": 0.0487, "step": 14753 }, { "epoch": 4.58, "learning_rate": 4.102020404080817e-06, "loss": 0.0103, "step": 14754 }, { "epoch": 4.58, "learning_rate": 4.1016203240648136e-06, "loss": 0.0228, "step": 14755 }, { "epoch": 4.58, "learning_rate": 4.10122024404881e-06, "loss": 0.0086, "step": 14756 }, { "epoch": 4.58, "learning_rate": 4.100820164032807e-06, "loss": 0.0414, "step": 14757 }, { "epoch": 4.58, "learning_rate": 4.100420084016804e-06, "loss": 0.0034, "step": 14758 }, { "epoch": 4.58, "learning_rate": 4.1000200040008e-06, "loss": 0.0112, "step": 14759 }, { "epoch": 4.58, "learning_rate": 4.099619923984797e-06, "loss": 0.0164, "step": 14760 }, { "epoch": 4.58, "learning_rate": 4.0992198439687945e-06, "loss": 0.0255, "step": 14761 }, { "epoch": 4.58, "learning_rate": 4.098819763952791e-06, "loss": 0.0166, "step": 14762 }, { "epoch": 4.58, "learning_rate": 4.098419683936788e-06, "loss": 0.0992, "step": 14763 }, { "epoch": 4.58, "learning_rate": 4.0980196039207846e-06, "loss": 0.0896, "step": 14764 }, { "epoch": 4.58, "learning_rate": 4.097619523904781e-06, "loss": 0.0314, "step": 14765 }, { "epoch": 4.58, "learning_rate": 4.097219443888778e-06, "loss": 0.0151, "step": 14766 }, { "epoch": 4.58, "learning_rate": 4.096819363872775e-06, "loss": 0.0118, "step": 14767 }, { "epoch": 4.58, "learning_rate": 4.096419283856772e-06, "loss": 0.0233, "step": 14768 }, { "epoch": 4.58, "learning_rate": 4.096019203840769e-06, "loss": 0.0251, "step": 14769 }, { "epoch": 4.58, "learning_rate": 4.0956191238247655e-06, "loss": 0.0118, "step": 14770 }, { "epoch": 4.58, "learning_rate": 4.095219043808762e-06, "loss": 0.0143, "step": 14771 }, { "epoch": 4.58, "learning_rate": 4.094818963792759e-06, "loss": 0.0391, "step": 14772 }, { "epoch": 4.59, "learning_rate": 4.094418883776756e-06, "loss": 0.0248, "step": 14773 }, { "epoch": 4.59, "learning_rate": 4.094018803760752e-06, "loss": 0.0029, "step": 14774 }, { "epoch": 4.59, "learning_rate": 4.09361872374475e-06, "loss": 0.0261, "step": 14775 }, { "epoch": 4.59, "learning_rate": 4.0932186437287465e-06, "loss": 0.0235, "step": 14776 }, { "epoch": 4.59, "learning_rate": 4.092818563712742e-06, "loss": 0.0139, "step": 14777 }, { "epoch": 4.59, "learning_rate": 4.09241848369674e-06, "loss": 0.0192, "step": 14778 }, { "epoch": 4.59, "learning_rate": 4.0920184036807366e-06, "loss": 0.0143, "step": 14779 }, { "epoch": 4.59, "learning_rate": 4.091618323664733e-06, "loss": 0.0171, "step": 14780 }, { "epoch": 4.59, "learning_rate": 4.09121824364873e-06, "loss": 0.0326, "step": 14781 }, { "epoch": 4.59, "learning_rate": 4.0908181636327275e-06, "loss": 0.0188, "step": 14782 }, { "epoch": 4.59, "learning_rate": 4.090418083616723e-06, "loss": 0.0196, "step": 14783 }, { "epoch": 4.59, "learning_rate": 4.09001800360072e-06, "loss": 0.0792, "step": 14784 }, { "epoch": 4.59, "learning_rate": 4.0896179235847175e-06, "loss": 0.0215, "step": 14785 }, { "epoch": 4.59, "learning_rate": 4.089217843568714e-06, "loss": 0.0125, "step": 14786 }, { "epoch": 4.59, "learning_rate": 4.088817763552711e-06, "loss": 0.0099, "step": 14787 }, { "epoch": 4.59, "learning_rate": 4.088417683536708e-06, "loss": 0.0548, "step": 14788 }, { "epoch": 4.59, "learning_rate": 4.088017603520704e-06, "loss": 0.0361, "step": 14789 }, { "epoch": 4.59, "learning_rate": 4.087617523504701e-06, "loss": 0.015, "step": 14790 }, { "epoch": 4.59, "learning_rate": 4.087217443488698e-06, "loss": 0.0175, "step": 14791 }, { "epoch": 4.59, "learning_rate": 4.086817363472695e-06, "loss": 0.012, "step": 14792 }, { "epoch": 4.59, "learning_rate": 4.086417283456692e-06, "loss": 0.0246, "step": 14793 }, { "epoch": 4.59, "learning_rate": 4.0860172034406886e-06, "loss": 0.0291, "step": 14794 }, { "epoch": 4.59, "learning_rate": 4.085617123424685e-06, "loss": 0.0269, "step": 14795 }, { "epoch": 4.59, "learning_rate": 4.085217043408682e-06, "loss": 0.0294, "step": 14796 }, { "epoch": 4.59, "learning_rate": 4.084816963392679e-06, "loss": 0.0223, "step": 14797 }, { "epoch": 4.59, "learning_rate": 4.084416883376675e-06, "loss": 0.0258, "step": 14798 }, { "epoch": 4.59, "learning_rate": 4.084016803360673e-06, "loss": 0.0216, "step": 14799 }, { "epoch": 4.59, "learning_rate": 4.0836167233446695e-06, "loss": 0.0179, "step": 14800 }, { "epoch": 4.59, "learning_rate": 4.083216643328666e-06, "loss": 0.0853, "step": 14801 }, { "epoch": 4.59, "learning_rate": 4.082816563312663e-06, "loss": 0.0385, "step": 14802 }, { "epoch": 4.59, "learning_rate": 4.0824164832966596e-06, "loss": 0.0301, "step": 14803 }, { "epoch": 4.59, "learning_rate": 4.082016403280656e-06, "loss": 0.0286, "step": 14804 }, { "epoch": 4.6, "learning_rate": 4.081616323264653e-06, "loss": 0.0235, "step": 14805 }, { "epoch": 4.6, "learning_rate": 4.0812162432486505e-06, "loss": 0.0103, "step": 14806 }, { "epoch": 4.6, "learning_rate": 4.080816163232647e-06, "loss": 0.0166, "step": 14807 }, { "epoch": 4.6, "learning_rate": 4.080416083216643e-06, "loss": 0.0358, "step": 14808 }, { "epoch": 4.6, "learning_rate": 4.0800160032006405e-06, "loss": 0.0343, "step": 14809 }, { "epoch": 4.6, "learning_rate": 4.079615923184637e-06, "loss": 0.0102, "step": 14810 }, { "epoch": 4.6, "learning_rate": 4.079215843168634e-06, "loss": 0.0168, "step": 14811 }, { "epoch": 4.6, "learning_rate": 4.078815763152631e-06, "loss": 0.0317, "step": 14812 }, { "epoch": 4.6, "learning_rate": 4.078415683136628e-06, "loss": 0.0355, "step": 14813 }, { "epoch": 4.6, "learning_rate": 4.078015603120625e-06, "loss": 0.0662, "step": 14814 }, { "epoch": 4.6, "learning_rate": 4.077615523104621e-06, "loss": 0.0271, "step": 14815 }, { "epoch": 4.6, "learning_rate": 4.077215443088618e-06, "loss": 0.0117, "step": 14816 }, { "epoch": 4.6, "learning_rate": 4.076815363072615e-06, "loss": 0.0104, "step": 14817 }, { "epoch": 4.6, "learning_rate": 4.0764152830566116e-06, "loss": 0.018, "step": 14818 }, { "epoch": 4.6, "learning_rate": 4.076015203040608e-06, "loss": 0.0264, "step": 14819 }, { "epoch": 4.6, "learning_rate": 4.075615123024606e-06, "loss": 0.0056, "step": 14820 }, { "epoch": 4.6, "learning_rate": 4.075215043008602e-06, "loss": 0.0108, "step": 14821 }, { "epoch": 4.6, "learning_rate": 4.074814962992598e-06, "loss": 0.0076, "step": 14822 }, { "epoch": 4.6, "learning_rate": 4.074414882976596e-06, "loss": 0.0198, "step": 14823 }, { "epoch": 4.6, "learning_rate": 4.0740148029605925e-06, "loss": 0.0296, "step": 14824 }, { "epoch": 4.6, "learning_rate": 4.073614722944589e-06, "loss": 0.0109, "step": 14825 }, { "epoch": 4.6, "learning_rate": 4.073214642928586e-06, "loss": 0.0146, "step": 14826 }, { "epoch": 4.6, "learning_rate": 4.072814562912583e-06, "loss": 0.0428, "step": 14827 }, { "epoch": 4.6, "learning_rate": 4.072414482896579e-06, "loss": 0.0194, "step": 14828 }, { "epoch": 4.6, "learning_rate": 4.072014402880576e-06, "loss": 0.0197, "step": 14829 }, { "epoch": 4.6, "learning_rate": 4.0716143228645735e-06, "loss": 0.0194, "step": 14830 }, { "epoch": 4.6, "learning_rate": 4.07121424284857e-06, "loss": 0.0092, "step": 14831 }, { "epoch": 4.6, "learning_rate": 4.070814162832567e-06, "loss": 0.0179, "step": 14832 }, { "epoch": 4.6, "learning_rate": 4.0704140828165636e-06, "loss": 0.0069, "step": 14833 }, { "epoch": 4.6, "learning_rate": 4.07001400280056e-06, "loss": 0.0226, "step": 14834 }, { "epoch": 4.6, "learning_rate": 4.069613922784557e-06, "loss": 0.0207, "step": 14835 }, { "epoch": 4.6, "learning_rate": 4.069213842768554e-06, "loss": 0.0201, "step": 14836 }, { "epoch": 4.61, "learning_rate": 4.068813762752551e-06, "loss": 0.0117, "step": 14837 }, { "epoch": 4.61, "learning_rate": 4.068413682736548e-06, "loss": 0.034, "step": 14838 }, { "epoch": 4.61, "learning_rate": 4.0680136027205445e-06, "loss": 0.0122, "step": 14839 }, { "epoch": 4.61, "learning_rate": 4.067613522704541e-06, "loss": 0.0115, "step": 14840 }, { "epoch": 4.61, "learning_rate": 4.067213442688538e-06, "loss": 0.029, "step": 14841 }, { "epoch": 4.61, "learning_rate": 4.0668133626725346e-06, "loss": 0.0162, "step": 14842 }, { "epoch": 4.61, "learning_rate": 4.066413282656531e-06, "loss": 0.0131, "step": 14843 }, { "epoch": 4.61, "learning_rate": 4.066013202640529e-06, "loss": 0.0181, "step": 14844 }, { "epoch": 4.61, "learning_rate": 4.0656131226245255e-06, "loss": 0.0075, "step": 14845 }, { "epoch": 4.61, "learning_rate": 4.065213042608522e-06, "loss": 0.0375, "step": 14846 }, { "epoch": 4.61, "learning_rate": 4.064812962592519e-06, "loss": 0.0482, "step": 14847 }, { "epoch": 4.61, "learning_rate": 4.0644128825765155e-06, "loss": 0.0218, "step": 14848 }, { "epoch": 4.61, "learning_rate": 4.064012802560512e-06, "loss": 0.0218, "step": 14849 }, { "epoch": 4.61, "learning_rate": 4.063612722544509e-06, "loss": 0.0121, "step": 14850 }, { "epoch": 4.61, "learning_rate": 4.0632126425285064e-06, "loss": 0.0185, "step": 14851 }, { "epoch": 4.61, "learning_rate": 4.062812562512502e-06, "loss": 0.0189, "step": 14852 }, { "epoch": 4.61, "learning_rate": 4.0624124824965e-06, "loss": 0.0104, "step": 14853 }, { "epoch": 4.61, "learning_rate": 4.0620124024804965e-06, "loss": 0.0231, "step": 14854 }, { "epoch": 4.61, "learning_rate": 4.061612322464493e-06, "loss": 0.0407, "step": 14855 }, { "epoch": 4.61, "learning_rate": 4.06121224244849e-06, "loss": 0.0161, "step": 14856 }, { "epoch": 4.61, "learning_rate": 4.060812162432487e-06, "loss": 0.0102, "step": 14857 }, { "epoch": 4.61, "learning_rate": 4.060412082416484e-06, "loss": 0.0068, "step": 14858 }, { "epoch": 4.61, "learning_rate": 4.06001200240048e-06, "loss": 0.0272, "step": 14859 }, { "epoch": 4.61, "learning_rate": 4.0596119223844775e-06, "loss": 0.0127, "step": 14860 }, { "epoch": 4.61, "learning_rate": 4.059211842368474e-06, "loss": 0.0454, "step": 14861 }, { "epoch": 4.61, "learning_rate": 4.058811762352471e-06, "loss": 0.0706, "step": 14862 }, { "epoch": 4.61, "learning_rate": 4.0584116823364675e-06, "loss": 0.019, "step": 14863 }, { "epoch": 4.61, "learning_rate": 4.058011602320465e-06, "loss": 0.0222, "step": 14864 }, { "epoch": 4.61, "learning_rate": 4.057611522304461e-06, "loss": 0.0447, "step": 14865 }, { "epoch": 4.61, "learning_rate": 4.057211442288458e-06, "loss": 0.0294, "step": 14866 }, { "epoch": 4.61, "learning_rate": 4.056811362272455e-06, "loss": 0.0176, "step": 14867 }, { "epoch": 4.61, "learning_rate": 4.056411282256452e-06, "loss": 0.0224, "step": 14868 }, { "epoch": 4.62, "learning_rate": 4.0560112022404485e-06, "loss": 0.0208, "step": 14869 }, { "epoch": 4.62, "learning_rate": 4.055611122224445e-06, "loss": 0.0082, "step": 14870 }, { "epoch": 4.62, "learning_rate": 4.055211042208442e-06, "loss": 0.0152, "step": 14871 }, { "epoch": 4.62, "learning_rate": 4.0548109621924386e-06, "loss": 0.1353, "step": 14872 }, { "epoch": 4.62, "learning_rate": 4.054410882176435e-06, "loss": 0.0302, "step": 14873 }, { "epoch": 4.62, "learning_rate": 4.054010802160433e-06, "loss": 0.0134, "step": 14874 }, { "epoch": 4.62, "learning_rate": 4.0536107221444295e-06, "loss": 0.0214, "step": 14875 }, { "epoch": 4.62, "learning_rate": 4.053210642128426e-06, "loss": 0.0216, "step": 14876 }, { "epoch": 4.62, "learning_rate": 4.052810562112423e-06, "loss": 0.0179, "step": 14877 }, { "epoch": 4.62, "learning_rate": 4.0524104820964195e-06, "loss": 0.0161, "step": 14878 }, { "epoch": 4.62, "learning_rate": 4.052010402080416e-06, "loss": 0.0228, "step": 14879 }, { "epoch": 4.62, "learning_rate": 4.051610322064413e-06, "loss": 0.0876, "step": 14880 }, { "epoch": 4.62, "learning_rate": 4.05121024204841e-06, "loss": 0.0087, "step": 14881 }, { "epoch": 4.62, "learning_rate": 4.050810162032407e-06, "loss": 0.0178, "step": 14882 }, { "epoch": 4.62, "learning_rate": 4.050410082016404e-06, "loss": 0.014, "step": 14883 }, { "epoch": 4.62, "learning_rate": 4.0500100020004005e-06, "loss": 0.0218, "step": 14884 }, { "epoch": 4.62, "learning_rate": 4.049609921984397e-06, "loss": 0.021, "step": 14885 }, { "epoch": 4.62, "learning_rate": 4.049209841968394e-06, "loss": 0.0078, "step": 14886 }, { "epoch": 4.62, "learning_rate": 4.0488097619523905e-06, "loss": 0.0234, "step": 14887 }, { "epoch": 4.62, "learning_rate": 4.048409681936388e-06, "loss": 0.0372, "step": 14888 }, { "epoch": 4.62, "learning_rate": 4.048009601920385e-06, "loss": 0.0141, "step": 14889 }, { "epoch": 4.62, "learning_rate": 4.047609521904381e-06, "loss": 0.0338, "step": 14890 }, { "epoch": 4.62, "learning_rate": 4.047209441888378e-06, "loss": 0.0069, "step": 14891 }, { "epoch": 4.62, "learning_rate": 4.046809361872375e-06, "loss": 0.0047, "step": 14892 }, { "epoch": 4.62, "learning_rate": 4.0464092818563715e-06, "loss": 0.0347, "step": 14893 }, { "epoch": 4.62, "learning_rate": 4.046009201840368e-06, "loss": 0.0348, "step": 14894 }, { "epoch": 4.62, "learning_rate": 4.045609121824366e-06, "loss": 0.0127, "step": 14895 }, { "epoch": 4.62, "learning_rate": 4.045209041808362e-06, "loss": 0.0244, "step": 14896 }, { "epoch": 4.62, "learning_rate": 4.044808961792358e-06, "loss": 0.0139, "step": 14897 }, { "epoch": 4.62, "learning_rate": 4.044408881776356e-06, "loss": 0.0112, "step": 14898 }, { "epoch": 4.62, "learning_rate": 4.0440088017603525e-06, "loss": 0.1023, "step": 14899 }, { "epoch": 4.62, "learning_rate": 4.043608721744349e-06, "loss": 0.0091, "step": 14900 }, { "epoch": 4.62, "learning_rate": 4.043208641728346e-06, "loss": 0.0262, "step": 14901 }, { "epoch": 4.63, "learning_rate": 4.042808561712343e-06, "loss": 0.065, "step": 14902 }, { "epoch": 4.63, "learning_rate": 4.042408481696339e-06, "loss": 0.0276, "step": 14903 }, { "epoch": 4.63, "learning_rate": 4.042008401680336e-06, "loss": 0.014, "step": 14904 }, { "epoch": 4.63, "learning_rate": 4.0416083216643334e-06, "loss": 0.0302, "step": 14905 }, { "epoch": 4.63, "learning_rate": 4.04120824164833e-06, "loss": 0.0312, "step": 14906 }, { "epoch": 4.63, "learning_rate": 4.040808161632327e-06, "loss": 0.0653, "step": 14907 }, { "epoch": 4.63, "learning_rate": 4.0404080816163235e-06, "loss": 0.016, "step": 14908 }, { "epoch": 4.63, "learning_rate": 4.04000800160032e-06, "loss": 0.0304, "step": 14909 }, { "epoch": 4.63, "learning_rate": 4.039607921584317e-06, "loss": 0.0249, "step": 14910 }, { "epoch": 4.63, "learning_rate": 4.0392078415683136e-06, "loss": 0.0719, "step": 14911 }, { "epoch": 4.63, "learning_rate": 4.038807761552311e-06, "loss": 0.0917, "step": 14912 }, { "epoch": 4.63, "learning_rate": 4.038407681536308e-06, "loss": 0.0209, "step": 14913 }, { "epoch": 4.63, "learning_rate": 4.0380076015203045e-06, "loss": 0.004, "step": 14914 }, { "epoch": 4.63, "learning_rate": 4.037607521504301e-06, "loss": 0.0084, "step": 14915 }, { "epoch": 4.63, "learning_rate": 4.037207441488298e-06, "loss": 0.0103, "step": 14916 }, { "epoch": 4.63, "learning_rate": 4.0368073614722945e-06, "loss": 0.0145, "step": 14917 }, { "epoch": 4.63, "learning_rate": 4.036407281456291e-06, "loss": 0.0059, "step": 14918 }, { "epoch": 4.63, "learning_rate": 4.036007201440289e-06, "loss": 0.0046, "step": 14919 }, { "epoch": 4.63, "learning_rate": 4.035607121424285e-06, "loss": 0.0049, "step": 14920 }, { "epoch": 4.63, "learning_rate": 4.035207041408282e-06, "loss": 0.0142, "step": 14921 }, { "epoch": 4.63, "learning_rate": 4.034806961392279e-06, "loss": 0.016, "step": 14922 }, { "epoch": 4.63, "learning_rate": 4.0344068813762755e-06, "loss": 0.0138, "step": 14923 }, { "epoch": 4.63, "learning_rate": 4.034006801360272e-06, "loss": 0.0119, "step": 14924 }, { "epoch": 4.63, "learning_rate": 4.033606721344269e-06, "loss": 0.0043, "step": 14925 }, { "epoch": 4.63, "learning_rate": 4.033206641328266e-06, "loss": 0.0105, "step": 14926 }, { "epoch": 4.63, "learning_rate": 4.032806561312263e-06, "loss": 0.0103, "step": 14927 }, { "epoch": 4.63, "learning_rate": 4.032406481296259e-06, "loss": 0.0192, "step": 14928 }, { "epoch": 4.63, "learning_rate": 4.0320064012802564e-06, "loss": 0.138, "step": 14929 }, { "epoch": 4.63, "learning_rate": 4.031606321264253e-06, "loss": 0.0374, "step": 14930 }, { "epoch": 4.63, "learning_rate": 4.03120624124825e-06, "loss": 0.0137, "step": 14931 }, { "epoch": 4.63, "learning_rate": 4.0308061612322465e-06, "loss": 0.0273, "step": 14932 }, { "epoch": 4.63, "learning_rate": 4.030406081216244e-06, "loss": 0.0203, "step": 14933 }, { "epoch": 4.64, "learning_rate": 4.03000600120024e-06, "loss": 0.0298, "step": 14934 }, { "epoch": 4.64, "learning_rate": 4.029605921184237e-06, "loss": 0.0174, "step": 14935 }, { "epoch": 4.64, "learning_rate": 4.029205841168234e-06, "loss": 0.026, "step": 14936 }, { "epoch": 4.64, "learning_rate": 4.028805761152231e-06, "loss": 0.0295, "step": 14937 }, { "epoch": 4.64, "learning_rate": 4.0284056811362275e-06, "loss": 0.023, "step": 14938 }, { "epoch": 4.64, "learning_rate": 4.028005601120224e-06, "loss": 0.0245, "step": 14939 }, { "epoch": 4.64, "learning_rate": 4.027605521104222e-06, "loss": 0.0168, "step": 14940 }, { "epoch": 4.64, "learning_rate": 4.0272054410882175e-06, "loss": 0.0203, "step": 14941 }, { "epoch": 4.64, "learning_rate": 4.026805361072215e-06, "loss": 0.0242, "step": 14942 }, { "epoch": 4.64, "learning_rate": 4.026405281056212e-06, "loss": 0.0185, "step": 14943 }, { "epoch": 4.64, "learning_rate": 4.0260052010402084e-06, "loss": 0.0381, "step": 14944 }, { "epoch": 4.64, "learning_rate": 4.025605121024205e-06, "loss": 0.0077, "step": 14945 }, { "epoch": 4.64, "learning_rate": 4.025205041008202e-06, "loss": 0.0215, "step": 14946 }, { "epoch": 4.64, "learning_rate": 4.0248049609921985e-06, "loss": 0.0139, "step": 14947 }, { "epoch": 4.64, "learning_rate": 4.024404880976195e-06, "loss": 0.0192, "step": 14948 }, { "epoch": 4.64, "learning_rate": 4.024004800960193e-06, "loss": 0.0159, "step": 14949 }, { "epoch": 4.64, "learning_rate": 4.023604720944189e-06, "loss": 0.0199, "step": 14950 }, { "epoch": 4.64, "learning_rate": 4.023204640928186e-06, "loss": 0.0244, "step": 14951 }, { "epoch": 4.64, "learning_rate": 4.022804560912183e-06, "loss": 0.017, "step": 14952 }, { "epoch": 4.64, "learning_rate": 4.0224044808961795e-06, "loss": 0.0709, "step": 14953 }, { "epoch": 4.64, "learning_rate": 4.022004400880176e-06, "loss": 0.0117, "step": 14954 }, { "epoch": 4.64, "learning_rate": 4.021604320864173e-06, "loss": 0.032, "step": 14955 }, { "epoch": 4.64, "learning_rate": 4.02120424084817e-06, "loss": 0.0176, "step": 14956 }, { "epoch": 4.64, "learning_rate": 4.020804160832167e-06, "loss": 0.0157, "step": 14957 }, { "epoch": 4.64, "learning_rate": 4.020404080816164e-06, "loss": 0.0224, "step": 14958 }, { "epoch": 4.64, "learning_rate": 4.02000400080016e-06, "loss": 0.0104, "step": 14959 }, { "epoch": 4.64, "learning_rate": 4.019603920784157e-06, "loss": 0.0351, "step": 14960 }, { "epoch": 4.64, "learning_rate": 4.019203840768154e-06, "loss": 0.0099, "step": 14961 }, { "epoch": 4.64, "learning_rate": 4.0188037607521505e-06, "loss": 0.0114, "step": 14962 }, { "epoch": 4.64, "learning_rate": 4.018403680736148e-06, "loss": 0.0055, "step": 14963 }, { "epoch": 4.64, "learning_rate": 4.018003600720145e-06, "loss": 0.0357, "step": 14964 }, { "epoch": 4.64, "learning_rate": 4.017603520704141e-06, "loss": 0.0233, "step": 14965 }, { "epoch": 4.65, "learning_rate": 4.017203440688138e-06, "loss": 0.0477, "step": 14966 }, { "epoch": 4.65, "learning_rate": 4.016803360672135e-06, "loss": 0.0119, "step": 14967 }, { "epoch": 4.65, "learning_rate": 4.0164032806561314e-06, "loss": 0.0096, "step": 14968 }, { "epoch": 4.65, "learning_rate": 4.016003200640128e-06, "loss": 0.0342, "step": 14969 }, { "epoch": 4.65, "learning_rate": 4.015603120624126e-06, "loss": 0.0095, "step": 14970 }, { "epoch": 4.65, "learning_rate": 4.015203040608122e-06, "loss": 0.0113, "step": 14971 }, { "epoch": 4.65, "learning_rate": 4.014802960592118e-06, "loss": 0.0086, "step": 14972 }, { "epoch": 4.65, "learning_rate": 4.014402880576116e-06, "loss": 0.0134, "step": 14973 }, { "epoch": 4.65, "learning_rate": 4.014002800560112e-06, "loss": 0.0114, "step": 14974 }, { "epoch": 4.65, "learning_rate": 4.013602720544109e-06, "loss": 0.0104, "step": 14975 }, { "epoch": 4.65, "learning_rate": 4.013202640528106e-06, "loss": 0.0088, "step": 14976 }, { "epoch": 4.65, "learning_rate": 4.012802560512103e-06, "loss": 0.0133, "step": 14977 }, { "epoch": 4.65, "learning_rate": 4.0124024804961e-06, "loss": 0.0292, "step": 14978 }, { "epoch": 4.65, "learning_rate": 4.012002400480096e-06, "loss": 0.0311, "step": 14979 }, { "epoch": 4.65, "learning_rate": 4.011602320464093e-06, "loss": 0.0106, "step": 14980 }, { "epoch": 4.65, "learning_rate": 4.01120224044809e-06, "loss": 0.0234, "step": 14981 }, { "epoch": 4.65, "learning_rate": 4.010802160432087e-06, "loss": 0.0238, "step": 14982 }, { "epoch": 4.65, "learning_rate": 4.0104020804160834e-06, "loss": 0.027, "step": 14983 }, { "epoch": 4.65, "learning_rate": 4.010002000400081e-06, "loss": 0.1096, "step": 14984 }, { "epoch": 4.65, "learning_rate": 4.009601920384077e-06, "loss": 0.0439, "step": 14985 }, { "epoch": 4.65, "learning_rate": 4.0092018403680735e-06, "loss": 0.0904, "step": 14986 }, { "epoch": 4.65, "learning_rate": 4.008801760352071e-06, "loss": 0.0124, "step": 14987 }, { "epoch": 4.65, "learning_rate": 4.008401680336068e-06, "loss": 0.0217, "step": 14988 }, { "epoch": 4.65, "learning_rate": 4.008001600320064e-06, "loss": 0.0034, "step": 14989 }, { "epoch": 4.65, "learning_rate": 4.007601520304061e-06, "loss": 0.0075, "step": 14990 }, { "epoch": 4.65, "learning_rate": 4.007201440288058e-06, "loss": 0.0128, "step": 14991 }, { "epoch": 4.65, "learning_rate": 4.0068013602720545e-06, "loss": 0.0167, "step": 14992 }, { "epoch": 4.65, "learning_rate": 4.006401280256051e-06, "loss": 0.0305, "step": 14993 }, { "epoch": 4.65, "learning_rate": 4.006001200240049e-06, "loss": 0.0101, "step": 14994 }, { "epoch": 4.65, "learning_rate": 4.005601120224045e-06, "loss": 0.0096, "step": 14995 }, { "epoch": 4.65, "learning_rate": 4.005201040208042e-06, "loss": 0.0146, "step": 14996 }, { "epoch": 4.65, "learning_rate": 4.004800960192039e-06, "loss": 0.0138, "step": 14997 }, { "epoch": 4.66, "learning_rate": 4.004400880176035e-06, "loss": 0.0294, "step": 14998 }, { "epoch": 4.66, "learning_rate": 4.004000800160032e-06, "loss": 0.0603, "step": 14999 }, { "epoch": 4.66, "learning_rate": 4.003600720144029e-06, "loss": 0.0166, "step": 15000 }, { "epoch": 4.66, "learning_rate": 4.003200640128026e-06, "loss": 0.0089, "step": 15001 }, { "epoch": 4.66, "learning_rate": 4.002800560112023e-06, "loss": 0.0106, "step": 15002 }, { "epoch": 4.66, "learning_rate": 4.00240048009602e-06, "loss": 0.0106, "step": 15003 }, { "epoch": 4.66, "learning_rate": 4.002000400080016e-06, "loss": 0.0321, "step": 15004 }, { "epoch": 4.66, "learning_rate": 4.001600320064013e-06, "loss": 0.0236, "step": 15005 }, { "epoch": 4.66, "learning_rate": 4.00120024004801e-06, "loss": 0.071, "step": 15006 }, { "epoch": 4.66, "learning_rate": 4.0008001600320064e-06, "loss": 0.1204, "step": 15007 }, { "epoch": 4.66, "learning_rate": 4.000400080016004e-06, "loss": 0.0067, "step": 15008 }, { "epoch": 4.66, "learning_rate": 4.000000000000001e-06, "loss": 0.0081, "step": 15009 }, { "epoch": 4.66, "learning_rate": 3.9995999199839965e-06, "loss": 0.008, "step": 15010 }, { "epoch": 4.66, "learning_rate": 3.999199839967994e-06, "loss": 0.0294, "step": 15011 }, { "epoch": 4.66, "learning_rate": 3.998799759951991e-06, "loss": 0.0106, "step": 15012 }, { "epoch": 4.66, "learning_rate": 3.998399679935987e-06, "loss": 0.0136, "step": 15013 }, { "epoch": 4.66, "learning_rate": 3.997999599919984e-06, "loss": 0.0258, "step": 15014 }, { "epoch": 4.66, "learning_rate": 3.997599519903982e-06, "loss": 0.0325, "step": 15015 }, { "epoch": 4.66, "learning_rate": 3.9971994398879775e-06, "loss": 0.0079, "step": 15016 }, { "epoch": 4.66, "learning_rate": 3.996799359871974e-06, "loss": 0.0199, "step": 15017 }, { "epoch": 4.66, "learning_rate": 3.996399279855972e-06, "loss": 0.0345, "step": 15018 }, { "epoch": 4.66, "learning_rate": 3.995999199839968e-06, "loss": 0.0231, "step": 15019 }, { "epoch": 4.66, "learning_rate": 3.995599119823965e-06, "loss": 0.0198, "step": 15020 }, { "epoch": 4.66, "learning_rate": 3.995199039807962e-06, "loss": 0.0472, "step": 15021 }, { "epoch": 4.66, "learning_rate": 3.994798959791959e-06, "loss": 0.0213, "step": 15022 }, { "epoch": 4.66, "learning_rate": 3.994398879775955e-06, "loss": 0.0232, "step": 15023 }, { "epoch": 4.66, "learning_rate": 3.993998799759952e-06, "loss": 0.0138, "step": 15024 }, { "epoch": 4.66, "learning_rate": 3.993598719743949e-06, "loss": 0.0625, "step": 15025 }, { "epoch": 4.66, "learning_rate": 3.993198639727946e-06, "loss": 0.009, "step": 15026 }, { "epoch": 4.66, "learning_rate": 3.992798559711943e-06, "loss": 0.0272, "step": 15027 }, { "epoch": 4.66, "learning_rate": 3.992398479695939e-06, "loss": 0.0229, "step": 15028 }, { "epoch": 4.66, "learning_rate": 3.991998399679936e-06, "loss": 0.0336, "step": 15029 }, { "epoch": 4.66, "learning_rate": 3.991598319663933e-06, "loss": 0.014, "step": 15030 }, { "epoch": 4.67, "learning_rate": 3.9911982396479295e-06, "loss": 0.0246, "step": 15031 }, { "epoch": 4.67, "learning_rate": 3.990798159631927e-06, "loss": 0.009, "step": 15032 }, { "epoch": 4.67, "learning_rate": 3.990398079615924e-06, "loss": 0.0529, "step": 15033 }, { "epoch": 4.67, "learning_rate": 3.98999799959992e-06, "loss": 0.0437, "step": 15034 }, { "epoch": 4.67, "learning_rate": 3.989597919583917e-06, "loss": 0.016, "step": 15035 }, { "epoch": 4.67, "learning_rate": 3.989197839567914e-06, "loss": 0.0106, "step": 15036 }, { "epoch": 4.67, "learning_rate": 3.98879775955191e-06, "loss": 0.0245, "step": 15037 }, { "epoch": 4.67, "learning_rate": 3.988397679535907e-06, "loss": 0.0149, "step": 15038 }, { "epoch": 4.67, "learning_rate": 3.987997599519905e-06, "loss": 0.0104, "step": 15039 }, { "epoch": 4.67, "learning_rate": 3.987597519503901e-06, "loss": 0.03, "step": 15040 }, { "epoch": 4.67, "learning_rate": 3.987197439487898e-06, "loss": 0.0222, "step": 15041 }, { "epoch": 4.67, "learning_rate": 3.986797359471895e-06, "loss": 0.0797, "step": 15042 }, { "epoch": 4.67, "learning_rate": 3.986397279455891e-06, "loss": 0.0804, "step": 15043 }, { "epoch": 4.67, "learning_rate": 3.985997199439888e-06, "loss": 0.0136, "step": 15044 }, { "epoch": 4.67, "learning_rate": 3.985597119423885e-06, "loss": 0.0202, "step": 15045 }, { "epoch": 4.67, "learning_rate": 3.985197039407882e-06, "loss": 0.0104, "step": 15046 }, { "epoch": 4.67, "learning_rate": 3.984796959391879e-06, "loss": 0.0026, "step": 15047 }, { "epoch": 4.67, "learning_rate": 3.984396879375876e-06, "loss": 0.0768, "step": 15048 }, { "epoch": 4.67, "learning_rate": 3.983996799359872e-06, "loss": 0.0265, "step": 15049 }, { "epoch": 4.67, "learning_rate": 3.983596719343869e-06, "loss": 0.0195, "step": 15050 }, { "epoch": 4.67, "learning_rate": 3.983196639327866e-06, "loss": 0.0111, "step": 15051 }, { "epoch": 4.67, "learning_rate": 3.982796559311862e-06, "loss": 0.0781, "step": 15052 }, { "epoch": 4.67, "learning_rate": 3.98239647929586e-06, "loss": 0.0318, "step": 15053 }, { "epoch": 4.67, "learning_rate": 3.981996399279856e-06, "loss": 0.0032, "step": 15054 }, { "epoch": 4.67, "learning_rate": 3.981596319263853e-06, "loss": 0.0068, "step": 15055 }, { "epoch": 4.67, "learning_rate": 3.98119623924785e-06, "loss": 0.0848, "step": 15056 }, { "epoch": 4.67, "learning_rate": 3.980796159231847e-06, "loss": 0.0365, "step": 15057 }, { "epoch": 4.67, "learning_rate": 3.980396079215843e-06, "loss": 0.0068, "step": 15058 }, { "epoch": 4.67, "learning_rate": 3.97999599919984e-06, "loss": 0.0142, "step": 15059 }, { "epoch": 4.67, "learning_rate": 3.979595919183838e-06, "loss": 0.0085, "step": 15060 }, { "epoch": 4.67, "learning_rate": 3.9791958391678334e-06, "loss": 0.0067, "step": 15061 }, { "epoch": 4.67, "learning_rate": 3.978795759151831e-06, "loss": 0.0883, "step": 15062 }, { "epoch": 4.68, "learning_rate": 3.978395679135828e-06, "loss": 0.0144, "step": 15063 }, { "epoch": 4.68, "learning_rate": 3.977995599119824e-06, "loss": 0.0285, "step": 15064 }, { "epoch": 4.68, "learning_rate": 3.977595519103821e-06, "loss": 0.0077, "step": 15065 }, { "epoch": 4.68, "learning_rate": 3.977195439087818e-06, "loss": 0.0148, "step": 15066 }, { "epoch": 4.68, "learning_rate": 3.976795359071814e-06, "loss": 0.0106, "step": 15067 }, { "epoch": 4.68, "learning_rate": 3.976395279055811e-06, "loss": 0.011, "step": 15068 }, { "epoch": 4.68, "learning_rate": 3.975995199039809e-06, "loss": 0.0204, "step": 15069 }, { "epoch": 4.68, "learning_rate": 3.975595119023805e-06, "loss": 0.0245, "step": 15070 }, { "epoch": 4.68, "learning_rate": 3.975195039007802e-06, "loss": 0.0249, "step": 15071 }, { "epoch": 4.68, "learning_rate": 3.974794958991799e-06, "loss": 0.0276, "step": 15072 }, { "epoch": 4.68, "learning_rate": 3.974394878975795e-06, "loss": 0.0111, "step": 15073 }, { "epoch": 4.68, "learning_rate": 3.973994798959792e-06, "loss": 0.0231, "step": 15074 }, { "epoch": 4.68, "learning_rate": 3.973594718943789e-06, "loss": 0.0189, "step": 15075 }, { "epoch": 4.68, "learning_rate": 3.973194638927786e-06, "loss": 0.0686, "step": 15076 }, { "epoch": 4.68, "learning_rate": 3.972794558911783e-06, "loss": 0.0187, "step": 15077 }, { "epoch": 4.68, "learning_rate": 3.97239447889578e-06, "loss": 0.1012, "step": 15078 }, { "epoch": 4.68, "learning_rate": 3.971994398879776e-06, "loss": 0.0139, "step": 15079 }, { "epoch": 4.68, "learning_rate": 3.971594318863773e-06, "loss": 0.0303, "step": 15080 }, { "epoch": 4.68, "learning_rate": 3.97119423884777e-06, "loss": 0.0279, "step": 15081 }, { "epoch": 4.68, "learning_rate": 3.970794158831766e-06, "loss": 0.0553, "step": 15082 }, { "epoch": 4.68, "learning_rate": 3.970394078815764e-06, "loss": 0.0174, "step": 15083 }, { "epoch": 4.68, "learning_rate": 3.969993998799761e-06, "loss": 0.0199, "step": 15084 }, { "epoch": 4.68, "learning_rate": 3.969593918783757e-06, "loss": 0.0175, "step": 15085 }, { "epoch": 4.68, "learning_rate": 3.969193838767754e-06, "loss": 0.0173, "step": 15086 }, { "epoch": 4.68, "learning_rate": 3.968793758751751e-06, "loss": 0.0164, "step": 15087 }, { "epoch": 4.68, "learning_rate": 3.968393678735747e-06, "loss": 0.0106, "step": 15088 }, { "epoch": 4.68, "learning_rate": 3.967993598719744e-06, "loss": 0.0374, "step": 15089 }, { "epoch": 4.68, "learning_rate": 3.9675935187037416e-06, "loss": 0.0381, "step": 15090 }, { "epoch": 4.68, "learning_rate": 3.967193438687738e-06, "loss": 0.0121, "step": 15091 }, { "epoch": 4.68, "learning_rate": 3.966793358671734e-06, "loss": 0.0146, "step": 15092 }, { "epoch": 4.68, "learning_rate": 3.966393278655732e-06, "loss": 0.0184, "step": 15093 }, { "epoch": 4.68, "learning_rate": 3.965993198639728e-06, "loss": 0.0134, "step": 15094 }, { "epoch": 4.69, "learning_rate": 3.965593118623725e-06, "loss": 0.0306, "step": 15095 }, { "epoch": 4.69, "learning_rate": 3.965193038607722e-06, "loss": 0.0204, "step": 15096 }, { "epoch": 4.69, "learning_rate": 3.964792958591719e-06, "loss": 0.0239, "step": 15097 }, { "epoch": 4.69, "learning_rate": 3.964392878575715e-06, "loss": 0.0199, "step": 15098 }, { "epoch": 4.69, "learning_rate": 3.963992798559712e-06, "loss": 0.0217, "step": 15099 }, { "epoch": 4.69, "learning_rate": 3.963592718543709e-06, "loss": 0.1144, "step": 15100 }, { "epoch": 4.69, "learning_rate": 3.963192638527706e-06, "loss": 0.0339, "step": 15101 }, { "epoch": 4.69, "learning_rate": 3.962792558511703e-06, "loss": 0.0675, "step": 15102 }, { "epoch": 4.69, "learning_rate": 3.962392478495699e-06, "loss": 0.0154, "step": 15103 }, { "epoch": 4.69, "learning_rate": 3.961992398479697e-06, "loss": 0.0131, "step": 15104 }, { "epoch": 4.69, "learning_rate": 3.961592318463693e-06, "loss": 0.0108, "step": 15105 }, { "epoch": 4.69, "learning_rate": 3.961192238447689e-06, "loss": 0.0295, "step": 15106 }, { "epoch": 4.69, "learning_rate": 3.960792158431687e-06, "loss": 0.0108, "step": 15107 }, { "epoch": 4.69, "learning_rate": 3.960392078415684e-06, "loss": 0.0213, "step": 15108 }, { "epoch": 4.69, "learning_rate": 3.95999199839968e-06, "loss": 0.0402, "step": 15109 }, { "epoch": 4.69, "learning_rate": 3.959591918383677e-06, "loss": 0.0072, "step": 15110 }, { "epoch": 4.69, "learning_rate": 3.959191838367674e-06, "loss": 0.0519, "step": 15111 }, { "epoch": 4.69, "learning_rate": 3.95879175835167e-06, "loss": 0.0115, "step": 15112 }, { "epoch": 4.69, "learning_rate": 3.958391678335667e-06, "loss": 0.0129, "step": 15113 }, { "epoch": 4.69, "learning_rate": 3.957991598319665e-06, "loss": 0.037, "step": 15114 }, { "epoch": 4.69, "learning_rate": 3.957591518303661e-06, "loss": 0.0275, "step": 15115 }, { "epoch": 4.69, "learning_rate": 3.957191438287658e-06, "loss": 0.0368, "step": 15116 }, { "epoch": 4.69, "learning_rate": 3.956791358271655e-06, "loss": 0.015, "step": 15117 }, { "epoch": 4.69, "learning_rate": 3.956391278255651e-06, "loss": 0.0084, "step": 15118 }, { "epoch": 4.69, "learning_rate": 3.955991198239648e-06, "loss": 0.0067, "step": 15119 }, { "epoch": 4.69, "learning_rate": 3.955591118223645e-06, "loss": 0.012, "step": 15120 }, { "epoch": 4.69, "learning_rate": 3.955191038207642e-06, "loss": 0.0316, "step": 15121 }, { "epoch": 4.69, "learning_rate": 3.954790958191639e-06, "loss": 0.0215, "step": 15122 }, { "epoch": 4.69, "learning_rate": 3.954390878175636e-06, "loss": 0.083, "step": 15123 }, { "epoch": 4.69, "learning_rate": 3.953990798159632e-06, "loss": 0.0146, "step": 15124 }, { "epoch": 4.69, "learning_rate": 3.953590718143629e-06, "loss": 0.0068, "step": 15125 }, { "epoch": 4.69, "learning_rate": 3.953190638127626e-06, "loss": 0.0189, "step": 15126 }, { "epoch": 4.7, "learning_rate": 3.952790558111622e-06, "loss": 0.0107, "step": 15127 }, { "epoch": 4.7, "learning_rate": 3.95239047809562e-06, "loss": 0.028, "step": 15128 }, { "epoch": 4.7, "learning_rate": 3.9519903980796166e-06, "loss": 0.0252, "step": 15129 }, { "epoch": 4.7, "learning_rate": 3.951590318063612e-06, "loss": 0.0146, "step": 15130 }, { "epoch": 4.7, "learning_rate": 3.95119023804761e-06, "loss": 0.0114, "step": 15131 }, { "epoch": 4.7, "learning_rate": 3.950790158031607e-06, "loss": 0.0127, "step": 15132 }, { "epoch": 4.7, "learning_rate": 3.950390078015603e-06, "loss": 0.0575, "step": 15133 }, { "epoch": 4.7, "learning_rate": 3.9499899979996e-06, "loss": 0.0945, "step": 15134 }, { "epoch": 4.7, "learning_rate": 3.9495899179835975e-06, "loss": 0.0238, "step": 15135 }, { "epoch": 4.7, "learning_rate": 3.949189837967593e-06, "loss": 0.0374, "step": 15136 }, { "epoch": 4.7, "learning_rate": 3.94878975795159e-06, "loss": 0.0122, "step": 15137 }, { "epoch": 4.7, "learning_rate": 3.948389677935588e-06, "loss": 0.0206, "step": 15138 }, { "epoch": 4.7, "learning_rate": 3.947989597919584e-06, "loss": 0.0116, "step": 15139 }, { "epoch": 4.7, "learning_rate": 3.947589517903581e-06, "loss": 0.0345, "step": 15140 }, { "epoch": 4.7, "learning_rate": 3.947189437887578e-06, "loss": 0.0142, "step": 15141 }, { "epoch": 4.7, "learning_rate": 3.946789357871575e-06, "loss": 0.0962, "step": 15142 }, { "epoch": 4.7, "learning_rate": 3.946389277855571e-06, "loss": 0.0142, "step": 15143 }, { "epoch": 4.7, "learning_rate": 3.945989197839568e-06, "loss": 0.0342, "step": 15144 }, { "epoch": 4.7, "learning_rate": 3.945589117823565e-06, "loss": 0.0345, "step": 15145 }, { "epoch": 4.7, "learning_rate": 3.945189037807562e-06, "loss": 0.0078, "step": 15146 }, { "epoch": 4.7, "learning_rate": 3.944788957791559e-06, "loss": 0.0077, "step": 15147 }, { "epoch": 4.7, "learning_rate": 3.944388877775555e-06, "loss": 0.0176, "step": 15148 }, { "epoch": 4.7, "learning_rate": 3.943988797759552e-06, "loss": 0.0184, "step": 15149 }, { "epoch": 4.7, "learning_rate": 3.943588717743549e-06, "loss": 0.0237, "step": 15150 }, { "epoch": 4.7, "learning_rate": 3.943188637727545e-06, "loss": 0.0075, "step": 15151 }, { "epoch": 4.7, "learning_rate": 3.942788557711543e-06, "loss": 0.0051, "step": 15152 }, { "epoch": 4.7, "learning_rate": 3.94238847769554e-06, "loss": 0.0041, "step": 15153 }, { "epoch": 4.7, "learning_rate": 3.941988397679536e-06, "loss": 0.1059, "step": 15154 }, { "epoch": 4.7, "learning_rate": 3.941588317663533e-06, "loss": 0.0265, "step": 15155 }, { "epoch": 4.7, "learning_rate": 3.94118823764753e-06, "loss": 0.0286, "step": 15156 }, { "epoch": 4.7, "learning_rate": 3.940788157631526e-06, "loss": 0.0112, "step": 15157 }, { "epoch": 4.7, "learning_rate": 3.940388077615524e-06, "loss": 0.0125, "step": 15158 }, { "epoch": 4.71, "learning_rate": 3.9399879975995205e-06, "loss": 0.0042, "step": 15159 }, { "epoch": 4.71, "learning_rate": 3.939587917583517e-06, "loss": 0.0068, "step": 15160 }, { "epoch": 4.71, "learning_rate": 3.939187837567514e-06, "loss": 0.0224, "step": 15161 }, { "epoch": 4.71, "learning_rate": 3.938787757551511e-06, "loss": 0.034, "step": 15162 }, { "epoch": 4.71, "learning_rate": 3.938387677535507e-06, "loss": 0.0059, "step": 15163 }, { "epoch": 4.71, "learning_rate": 3.937987597519504e-06, "loss": 0.013, "step": 15164 }, { "epoch": 4.71, "learning_rate": 3.9375875175035015e-06, "loss": 0.0134, "step": 15165 }, { "epoch": 4.71, "learning_rate": 3.937187437487498e-06, "loss": 0.0139, "step": 15166 }, { "epoch": 4.71, "learning_rate": 3.936787357471495e-06, "loss": 0.0362, "step": 15167 }, { "epoch": 4.71, "learning_rate": 3.9363872774554916e-06, "loss": 0.0525, "step": 15168 }, { "epoch": 4.71, "learning_rate": 3.935987197439488e-06, "loss": 0.008, "step": 15169 }, { "epoch": 4.71, "learning_rate": 3.935587117423485e-06, "loss": 0.0047, "step": 15170 }, { "epoch": 4.71, "learning_rate": 3.935187037407482e-06, "loss": 0.0291, "step": 15171 }, { "epoch": 4.71, "learning_rate": 3.934786957391479e-06, "loss": 0.0225, "step": 15172 }, { "epoch": 4.71, "learning_rate": 3.934386877375476e-06, "loss": 0.0084, "step": 15173 }, { "epoch": 4.71, "learning_rate": 3.933986797359472e-06, "loss": 0.115, "step": 15174 }, { "epoch": 4.71, "learning_rate": 3.933586717343469e-06, "loss": 0.0148, "step": 15175 }, { "epoch": 4.71, "learning_rate": 3.933186637327466e-06, "loss": 0.0099, "step": 15176 }, { "epoch": 4.71, "learning_rate": 3.932786557311463e-06, "loss": 0.0506, "step": 15177 }, { "epoch": 4.71, "learning_rate": 3.932386477295459e-06, "loss": 0.005, "step": 15178 }, { "epoch": 4.71, "learning_rate": 3.931986397279457e-06, "loss": 0.0208, "step": 15179 }, { "epoch": 4.71, "learning_rate": 3.931586317263453e-06, "loss": 0.0099, "step": 15180 }, { "epoch": 4.71, "learning_rate": 3.931186237247449e-06, "loss": 0.0323, "step": 15181 }, { "epoch": 4.71, "learning_rate": 3.930786157231447e-06, "loss": 0.0101, "step": 15182 }, { "epoch": 4.71, "learning_rate": 3.9303860772154436e-06, "loss": 0.0237, "step": 15183 }, { "epoch": 4.71, "learning_rate": 3.92998599719944e-06, "loss": 0.0197, "step": 15184 }, { "epoch": 4.71, "learning_rate": 3.929585917183437e-06, "loss": 0.0098, "step": 15185 }, { "epoch": 4.71, "learning_rate": 3.9291858371674345e-06, "loss": 0.0294, "step": 15186 }, { "epoch": 4.71, "learning_rate": 3.92878575715143e-06, "loss": 0.0059, "step": 15187 }, { "epoch": 4.71, "learning_rate": 3.928385677135427e-06, "loss": 0.0158, "step": 15188 }, { "epoch": 4.71, "learning_rate": 3.9279855971194245e-06, "loss": 0.1156, "step": 15189 }, { "epoch": 4.71, "learning_rate": 3.927585517103421e-06, "loss": 0.0549, "step": 15190 }, { "epoch": 4.71, "learning_rate": 3.927185437087418e-06, "loss": 0.0704, "step": 15191 }, { "epoch": 4.72, "learning_rate": 3.926785357071415e-06, "loss": 0.0541, "step": 15192 }, { "epoch": 4.72, "learning_rate": 3.926385277055411e-06, "loss": 0.0106, "step": 15193 }, { "epoch": 4.72, "learning_rate": 3.925985197039408e-06, "loss": 0.0873, "step": 15194 }, { "epoch": 4.72, "learning_rate": 3.925585117023405e-06, "loss": 0.0259, "step": 15195 }, { "epoch": 4.72, "learning_rate": 3.925185037007402e-06, "loss": 0.0357, "step": 15196 }, { "epoch": 4.72, "learning_rate": 3.924784956991399e-06, "loss": 0.0169, "step": 15197 }, { "epoch": 4.72, "learning_rate": 3.9243848769753955e-06, "loss": 0.0142, "step": 15198 }, { "epoch": 4.72, "learning_rate": 3.923984796959392e-06, "loss": 0.0152, "step": 15199 }, { "epoch": 4.72, "learning_rate": 3.923584716943389e-06, "loss": 0.0291, "step": 15200 }, { "epoch": 4.72, "learning_rate": 3.923184636927386e-06, "loss": 0.0198, "step": 15201 }, { "epoch": 4.72, "learning_rate": 3.922784556911382e-06, "loss": 0.0192, "step": 15202 }, { "epoch": 4.72, "learning_rate": 3.92238447689538e-06, "loss": 0.0258, "step": 15203 }, { "epoch": 4.72, "learning_rate": 3.9219843968793765e-06, "loss": 0.0231, "step": 15204 }, { "epoch": 4.72, "learning_rate": 3.921584316863373e-06, "loss": 0.0097, "step": 15205 }, { "epoch": 4.72, "learning_rate": 3.92118423684737e-06, "loss": 0.0104, "step": 15206 }, { "epoch": 4.72, "learning_rate": 3.9207841568313666e-06, "loss": 0.0208, "step": 15207 }, { "epoch": 4.72, "learning_rate": 3.920384076815363e-06, "loss": 0.0183, "step": 15208 }, { "epoch": 4.72, "learning_rate": 3.91998399679936e-06, "loss": 0.0271, "step": 15209 }, { "epoch": 4.72, "learning_rate": 3.9195839167833575e-06, "loss": 0.0544, "step": 15210 }, { "epoch": 4.72, "learning_rate": 3.919183836767354e-06, "loss": 0.0108, "step": 15211 }, { "epoch": 4.72, "learning_rate": 3.91878375675135e-06, "loss": 0.0187, "step": 15212 }, { "epoch": 4.72, "learning_rate": 3.9183836767353475e-06, "loss": 0.019, "step": 15213 }, { "epoch": 4.72, "learning_rate": 3.917983596719344e-06, "loss": 0.0495, "step": 15214 }, { "epoch": 4.72, "learning_rate": 3.917583516703341e-06, "loss": 0.025, "step": 15215 }, { "epoch": 4.72, "learning_rate": 3.917183436687338e-06, "loss": 0.0494, "step": 15216 }, { "epoch": 4.72, "learning_rate": 3.916783356671335e-06, "loss": 0.0537, "step": 15217 }, { "epoch": 4.72, "learning_rate": 3.916383276655331e-06, "loss": 0.0187, "step": 15218 }, { "epoch": 4.72, "learning_rate": 3.915983196639328e-06, "loss": 0.0093, "step": 15219 }, { "epoch": 4.72, "learning_rate": 3.915583116623325e-06, "loss": 0.0463, "step": 15220 }, { "epoch": 4.72, "learning_rate": 3.915183036607322e-06, "loss": 0.0352, "step": 15221 }, { "epoch": 4.72, "learning_rate": 3.9147829565913186e-06, "loss": 0.0315, "step": 15222 }, { "epoch": 4.72, "learning_rate": 3.914382876575315e-06, "loss": 0.0215, "step": 15223 }, { "epoch": 4.73, "learning_rate": 3.913982796559313e-06, "loss": 0.0097, "step": 15224 }, { "epoch": 4.73, "learning_rate": 3.913582716543309e-06, "loss": 0.0265, "step": 15225 }, { "epoch": 4.73, "learning_rate": 3.913182636527305e-06, "loss": 0.0256, "step": 15226 }, { "epoch": 4.73, "learning_rate": 3.912782556511303e-06, "loss": 0.017, "step": 15227 }, { "epoch": 4.73, "learning_rate": 3.9123824764952995e-06, "loss": 0.0184, "step": 15228 }, { "epoch": 4.73, "learning_rate": 3.911982396479296e-06, "loss": 0.0383, "step": 15229 }, { "epoch": 4.73, "learning_rate": 3.911582316463293e-06, "loss": 0.0227, "step": 15230 }, { "epoch": 4.73, "learning_rate": 3.91118223644729e-06, "loss": 0.0063, "step": 15231 }, { "epoch": 4.73, "learning_rate": 3.910782156431286e-06, "loss": 0.029, "step": 15232 }, { "epoch": 4.73, "learning_rate": 3.910382076415283e-06, "loss": 0.0459, "step": 15233 }, { "epoch": 4.73, "learning_rate": 3.9099819963992805e-06, "loss": 0.0103, "step": 15234 }, { "epoch": 4.73, "learning_rate": 3.909581916383277e-06, "loss": 0.0052, "step": 15235 }, { "epoch": 4.73, "learning_rate": 3.909181836367274e-06, "loss": 0.0185, "step": 15236 }, { "epoch": 4.73, "learning_rate": 3.9087817563512705e-06, "loss": 0.0078, "step": 15237 }, { "epoch": 4.73, "learning_rate": 3.908381676335267e-06, "loss": 0.0453, "step": 15238 }, { "epoch": 4.73, "learning_rate": 3.907981596319264e-06, "loss": 0.0097, "step": 15239 }, { "epoch": 4.73, "learning_rate": 3.907581516303261e-06, "loss": 0.0132, "step": 15240 }, { "epoch": 4.73, "learning_rate": 3.907181436287258e-06, "loss": 0.0248, "step": 15241 }, { "epoch": 4.73, "learning_rate": 3.906781356271255e-06, "loss": 0.0749, "step": 15242 }, { "epoch": 4.73, "learning_rate": 3.9063812762552515e-06, "loss": 0.0075, "step": 15243 }, { "epoch": 4.73, "learning_rate": 3.905981196239248e-06, "loss": 0.0068, "step": 15244 }, { "epoch": 4.73, "learning_rate": 3.905581116223245e-06, "loss": 0.0355, "step": 15245 }, { "epoch": 4.73, "learning_rate": 3.9051810362072416e-06, "loss": 0.013, "step": 15246 }, { "epoch": 4.73, "learning_rate": 3.904780956191238e-06, "loss": 0.0246, "step": 15247 }, { "epoch": 4.73, "learning_rate": 3.904380876175236e-06, "loss": 0.02, "step": 15248 }, { "epoch": 4.73, "learning_rate": 3.9039807961592325e-06, "loss": 0.0118, "step": 15249 }, { "epoch": 4.73, "learning_rate": 3.903580716143229e-06, "loss": 0.0249, "step": 15250 }, { "epoch": 4.73, "learning_rate": 3.903180636127226e-06, "loss": 0.0222, "step": 15251 }, { "epoch": 4.73, "learning_rate": 3.9027805561112225e-06, "loss": 0.0452, "step": 15252 }, { "epoch": 4.73, "learning_rate": 3.902380476095219e-06, "loss": 0.0183, "step": 15253 }, { "epoch": 4.73, "learning_rate": 3.901980396079216e-06, "loss": 0.0302, "step": 15254 }, { "epoch": 4.73, "learning_rate": 3.9015803160632134e-06, "loss": 0.0116, "step": 15255 }, { "epoch": 4.74, "learning_rate": 3.901180236047209e-06, "loss": 0.0379, "step": 15256 }, { "epoch": 4.74, "learning_rate": 3.900780156031207e-06, "loss": 0.0165, "step": 15257 }, { "epoch": 4.74, "learning_rate": 3.9003800760152035e-06, "loss": 0.023, "step": 15258 }, { "epoch": 4.74, "learning_rate": 3.8999799959992e-06, "loss": 0.023, "step": 15259 }, { "epoch": 4.74, "learning_rate": 3.899579915983197e-06, "loss": 0.0264, "step": 15260 }, { "epoch": 4.74, "learning_rate": 3.8991798359671936e-06, "loss": 0.0133, "step": 15261 }, { "epoch": 4.74, "learning_rate": 3.89877975595119e-06, "loss": 0.0202, "step": 15262 }, { "epoch": 4.74, "learning_rate": 3.898379675935187e-06, "loss": 0.014, "step": 15263 }, { "epoch": 4.74, "learning_rate": 3.8979795959191845e-06, "loss": 0.053, "step": 15264 }, { "epoch": 4.74, "learning_rate": 3.897579515903181e-06, "loss": 0.0075, "step": 15265 }, { "epoch": 4.74, "learning_rate": 3.897179435887178e-06, "loss": 0.0178, "step": 15266 }, { "epoch": 4.74, "learning_rate": 3.8967793558711745e-06, "loss": 0.0065, "step": 15267 }, { "epoch": 4.74, "learning_rate": 3.896379275855171e-06, "loss": 0.0106, "step": 15268 }, { "epoch": 4.74, "learning_rate": 3.895979195839168e-06, "loss": 0.0336, "step": 15269 }, { "epoch": 4.74, "learning_rate": 3.895579115823165e-06, "loss": 0.0178, "step": 15270 }, { "epoch": 4.74, "learning_rate": 3.895179035807162e-06, "loss": 0.0413, "step": 15271 }, { "epoch": 4.74, "learning_rate": 3.894778955791159e-06, "loss": 0.0268, "step": 15272 }, { "epoch": 4.74, "learning_rate": 3.8943788757751555e-06, "loss": 0.0075, "step": 15273 }, { "epoch": 4.74, "learning_rate": 3.893978795759152e-06, "loss": 0.0285, "step": 15274 }, { "epoch": 4.74, "learning_rate": 3.893578715743149e-06, "loss": 0.0903, "step": 15275 }, { "epoch": 4.74, "learning_rate": 3.8931786357271455e-06, "loss": 0.0194, "step": 15276 }, { "epoch": 4.74, "learning_rate": 3.892778555711142e-06, "loss": 0.0473, "step": 15277 }, { "epoch": 4.74, "learning_rate": 3.89237847569514e-06, "loss": 0.0239, "step": 15278 }, { "epoch": 4.74, "learning_rate": 3.8919783956791365e-06, "loss": 0.0232, "step": 15279 }, { "epoch": 4.74, "learning_rate": 3.891578315663133e-06, "loss": 0.0076, "step": 15280 }, { "epoch": 4.74, "learning_rate": 3.89117823564713e-06, "loss": 0.0065, "step": 15281 }, { "epoch": 4.74, "learning_rate": 3.8907781556311265e-06, "loss": 0.0316, "step": 15282 }, { "epoch": 4.74, "learning_rate": 3.890378075615123e-06, "loss": 0.0087, "step": 15283 }, { "epoch": 4.74, "learning_rate": 3.88997799559912e-06, "loss": 0.0204, "step": 15284 }, { "epoch": 4.74, "learning_rate": 3.889577915583117e-06, "loss": 0.0287, "step": 15285 }, { "epoch": 4.74, "learning_rate": 3.889177835567114e-06, "loss": 0.0239, "step": 15286 }, { "epoch": 4.74, "learning_rate": 3.888777755551111e-06, "loss": 0.0153, "step": 15287 }, { "epoch": 4.75, "learning_rate": 3.8883776755351075e-06, "loss": 0.0144, "step": 15288 }, { "epoch": 4.75, "learning_rate": 3.887977595519104e-06, "loss": 0.0077, "step": 15289 }, { "epoch": 4.75, "learning_rate": 3.887577515503101e-06, "loss": 0.0175, "step": 15290 }, { "epoch": 4.75, "learning_rate": 3.8871774354870975e-06, "loss": 0.0698, "step": 15291 }, { "epoch": 4.75, "learning_rate": 3.886777355471095e-06, "loss": 0.0279, "step": 15292 }, { "epoch": 4.75, "learning_rate": 3.886377275455092e-06, "loss": 0.0253, "step": 15293 }, { "epoch": 4.75, "learning_rate": 3.885977195439088e-06, "loss": 0.0178, "step": 15294 }, { "epoch": 4.75, "learning_rate": 3.885577115423085e-06, "loss": 0.0209, "step": 15295 }, { "epoch": 4.75, "learning_rate": 3.885177035407082e-06, "loss": 0.0334, "step": 15296 }, { "epoch": 4.75, "learning_rate": 3.8847769553910785e-06, "loss": 0.045, "step": 15297 }, { "epoch": 4.75, "learning_rate": 3.884376875375075e-06, "loss": 0.0217, "step": 15298 }, { "epoch": 4.75, "learning_rate": 3.883976795359073e-06, "loss": 0.0094, "step": 15299 }, { "epoch": 4.75, "learning_rate": 3.8835767153430686e-06, "loss": 0.0189, "step": 15300 }, { "epoch": 4.75, "learning_rate": 3.883176635327065e-06, "loss": 0.0183, "step": 15301 }, { "epoch": 4.75, "learning_rate": 3.882776555311063e-06, "loss": 0.0179, "step": 15302 }, { "epoch": 4.75, "learning_rate": 3.8823764752950595e-06, "loss": 0.017, "step": 15303 }, { "epoch": 4.75, "learning_rate": 3.881976395279056e-06, "loss": 0.0182, "step": 15304 }, { "epoch": 4.75, "learning_rate": 3.881576315263053e-06, "loss": 0.0094, "step": 15305 }, { "epoch": 4.75, "learning_rate": 3.8811762352470495e-06, "loss": 0.0112, "step": 15306 }, { "epoch": 4.75, "learning_rate": 3.880776155231046e-06, "loss": 0.0086, "step": 15307 }, { "epoch": 4.75, "learning_rate": 3.880376075215043e-06, "loss": 0.0263, "step": 15308 }, { "epoch": 4.75, "learning_rate": 3.8799759951990404e-06, "loss": 0.0809, "step": 15309 }, { "epoch": 4.75, "learning_rate": 3.879575915183037e-06, "loss": 0.0206, "step": 15310 }, { "epoch": 4.75, "learning_rate": 3.879175835167034e-06, "loss": 0.0057, "step": 15311 }, { "epoch": 4.75, "learning_rate": 3.8787757551510305e-06, "loss": 0.0156, "step": 15312 }, { "epoch": 4.75, "learning_rate": 3.878375675135027e-06, "loss": 0.0265, "step": 15313 }, { "epoch": 4.75, "learning_rate": 3.877975595119024e-06, "loss": 0.0063, "step": 15314 }, { "epoch": 4.75, "learning_rate": 3.8775755151030205e-06, "loss": 0.0062, "step": 15315 }, { "epoch": 4.75, "learning_rate": 3.877175435087018e-06, "loss": 0.0134, "step": 15316 }, { "epoch": 4.75, "learning_rate": 3.876775355071015e-06, "loss": 0.0259, "step": 15317 }, { "epoch": 4.75, "learning_rate": 3.8763752750550115e-06, "loss": 0.0523, "step": 15318 }, { "epoch": 4.75, "learning_rate": 3.875975195039008e-06, "loss": 0.028, "step": 15319 }, { "epoch": 4.75, "learning_rate": 3.875575115023005e-06, "loss": 0.0304, "step": 15320 }, { "epoch": 4.76, "learning_rate": 3.8751750350070015e-06, "loss": 0.0354, "step": 15321 }, { "epoch": 4.76, "learning_rate": 3.874774954990998e-06, "loss": 0.0163, "step": 15322 }, { "epoch": 4.76, "learning_rate": 3.874374874974996e-06, "loss": 0.0267, "step": 15323 }, { "epoch": 4.76, "learning_rate": 3.873974794958992e-06, "loss": 0.0173, "step": 15324 }, { "epoch": 4.76, "learning_rate": 3.873574714942988e-06, "loss": 0.022, "step": 15325 }, { "epoch": 4.76, "learning_rate": 3.873174634926986e-06, "loss": 0.0083, "step": 15326 }, { "epoch": 4.76, "learning_rate": 3.8727745549109825e-06, "loss": 0.0094, "step": 15327 }, { "epoch": 4.76, "learning_rate": 3.872374474894979e-06, "loss": 0.0092, "step": 15328 }, { "epoch": 4.76, "learning_rate": 3.871974394878976e-06, "loss": 0.0178, "step": 15329 }, { "epoch": 4.76, "learning_rate": 3.871574314862973e-06, "loss": 0.0401, "step": 15330 }, { "epoch": 4.76, "learning_rate": 3.87117423484697e-06, "loss": 0.0776, "step": 15331 }, { "epoch": 4.76, "learning_rate": 3.870774154830966e-06, "loss": 0.0056, "step": 15332 }, { "epoch": 4.76, "learning_rate": 3.8703740748149634e-06, "loss": 0.0366, "step": 15333 }, { "epoch": 4.76, "learning_rate": 3.86997399479896e-06, "loss": 0.0201, "step": 15334 }, { "epoch": 4.76, "learning_rate": 3.869573914782957e-06, "loss": 0.0358, "step": 15335 }, { "epoch": 4.76, "learning_rate": 3.8691738347669535e-06, "loss": 0.0171, "step": 15336 }, { "epoch": 4.76, "learning_rate": 3.868773754750951e-06, "loss": 0.0281, "step": 15337 }, { "epoch": 4.76, "learning_rate": 3.868373674734947e-06, "loss": 0.024, "step": 15338 }, { "epoch": 4.76, "learning_rate": 3.8679735947189436e-06, "loss": 0.0244, "step": 15339 }, { "epoch": 4.76, "learning_rate": 3.867573514702941e-06, "loss": 0.027, "step": 15340 }, { "epoch": 4.76, "learning_rate": 3.867173434686938e-06, "loss": 0.0201, "step": 15341 }, { "epoch": 4.76, "learning_rate": 3.8667733546709345e-06, "loss": 0.0227, "step": 15342 }, { "epoch": 4.76, "learning_rate": 3.866373274654931e-06, "loss": 0.0252, "step": 15343 }, { "epoch": 4.76, "learning_rate": 3.865973194638928e-06, "loss": 0.0212, "step": 15344 }, { "epoch": 4.76, "learning_rate": 3.8655731146229245e-06, "loss": 0.0147, "step": 15345 }, { "epoch": 4.76, "learning_rate": 3.865173034606921e-06, "loss": 0.0579, "step": 15346 }, { "epoch": 4.76, "learning_rate": 3.864772954590919e-06, "loss": 0.0099, "step": 15347 }, { "epoch": 4.76, "learning_rate": 3.8643728745749154e-06, "loss": 0.0071, "step": 15348 }, { "epoch": 4.76, "learning_rate": 3.863972794558912e-06, "loss": 0.017, "step": 15349 }, { "epoch": 4.76, "learning_rate": 3.863572714542909e-06, "loss": 0.0203, "step": 15350 }, { "epoch": 4.76, "learning_rate": 3.8631726345269055e-06, "loss": 0.0488, "step": 15351 }, { "epoch": 4.76, "learning_rate": 3.862772554510902e-06, "loss": 0.0116, "step": 15352 }, { "epoch": 4.77, "learning_rate": 3.862372474494899e-06, "loss": 0.0098, "step": 15353 }, { "epoch": 4.77, "learning_rate": 3.861972394478896e-06, "loss": 0.0341, "step": 15354 }, { "epoch": 4.77, "learning_rate": 3.861572314462893e-06, "loss": 0.0156, "step": 15355 }, { "epoch": 4.77, "learning_rate": 3.86117223444689e-06, "loss": 0.059, "step": 15356 }, { "epoch": 4.77, "learning_rate": 3.8607721544308865e-06, "loss": 0.0282, "step": 15357 }, { "epoch": 4.77, "learning_rate": 3.860372074414883e-06, "loss": 0.0085, "step": 15358 }, { "epoch": 4.77, "learning_rate": 3.85997199439888e-06, "loss": 0.0192, "step": 15359 }, { "epoch": 4.77, "learning_rate": 3.8595719143828765e-06, "loss": 0.0081, "step": 15360 }, { "epoch": 4.77, "learning_rate": 3.859171834366874e-06, "loss": 0.1023, "step": 15361 }, { "epoch": 4.77, "learning_rate": 3.858771754350871e-06, "loss": 0.0419, "step": 15362 }, { "epoch": 4.77, "learning_rate": 3.858371674334867e-06, "loss": 0.0218, "step": 15363 }, { "epoch": 4.77, "learning_rate": 3.857971594318864e-06, "loss": 0.0069, "step": 15364 }, { "epoch": 4.77, "learning_rate": 3.857571514302861e-06, "loss": 0.0096, "step": 15365 }, { "epoch": 4.77, "learning_rate": 3.8571714342868575e-06, "loss": 0.0151, "step": 15366 }, { "epoch": 4.77, "learning_rate": 3.856771354270854e-06, "loss": 0.0049, "step": 15367 }, { "epoch": 4.77, "learning_rate": 3.856371274254852e-06, "loss": 0.0243, "step": 15368 }, { "epoch": 4.77, "learning_rate": 3.855971194238848e-06, "loss": 0.0126, "step": 15369 }, { "epoch": 4.77, "learning_rate": 3.855571114222845e-06, "loss": 0.0062, "step": 15370 }, { "epoch": 4.77, "learning_rate": 3.855171034206842e-06, "loss": 0.0205, "step": 15371 }, { "epoch": 4.77, "learning_rate": 3.8547709541908384e-06, "loss": 0.0161, "step": 15372 }, { "epoch": 4.77, "learning_rate": 3.854370874174835e-06, "loss": 0.0193, "step": 15373 }, { "epoch": 4.77, "learning_rate": 3.853970794158833e-06, "loss": 0.0061, "step": 15374 }, { "epoch": 4.77, "learning_rate": 3.853570714142829e-06, "loss": 0.0185, "step": 15375 }, { "epoch": 4.77, "learning_rate": 3.853170634126825e-06, "loss": 0.0717, "step": 15376 }, { "epoch": 4.77, "learning_rate": 3.852770554110823e-06, "loss": 0.0131, "step": 15377 }, { "epoch": 4.77, "learning_rate": 3.852370474094819e-06, "loss": 0.0065, "step": 15378 }, { "epoch": 4.77, "learning_rate": 3.851970394078816e-06, "loss": 0.0088, "step": 15379 }, { "epoch": 4.77, "learning_rate": 3.851570314062813e-06, "loss": 0.012, "step": 15380 }, { "epoch": 4.77, "learning_rate": 3.85117023404681e-06, "loss": 0.0515, "step": 15381 }, { "epoch": 4.77, "learning_rate": 3.850770154030806e-06, "loss": 0.0241, "step": 15382 }, { "epoch": 4.77, "learning_rate": 3.850370074014803e-06, "loss": 0.0187, "step": 15383 }, { "epoch": 4.77, "learning_rate": 3.8499699939988e-06, "loss": 0.015, "step": 15384 }, { "epoch": 4.78, "learning_rate": 3.849569913982797e-06, "loss": 0.0072, "step": 15385 }, { "epoch": 4.78, "learning_rate": 3.849169833966794e-06, "loss": 0.0339, "step": 15386 }, { "epoch": 4.78, "learning_rate": 3.8487697539507904e-06, "loss": 0.0142, "step": 15387 }, { "epoch": 4.78, "learning_rate": 3.848369673934787e-06, "loss": 0.0979, "step": 15388 }, { "epoch": 4.78, "learning_rate": 3.847969593918784e-06, "loss": 0.0203, "step": 15389 }, { "epoch": 4.78, "learning_rate": 3.8475695139027805e-06, "loss": 0.0211, "step": 15390 }, { "epoch": 4.78, "learning_rate": 3.847169433886778e-06, "loss": 0.0253, "step": 15391 }, { "epoch": 4.78, "learning_rate": 3.846769353870775e-06, "loss": 0.0023, "step": 15392 }, { "epoch": 4.78, "learning_rate": 3.846369273854771e-06, "loss": 0.0137, "step": 15393 }, { "epoch": 4.78, "learning_rate": 3.845969193838768e-06, "loss": 0.0157, "step": 15394 }, { "epoch": 4.78, "learning_rate": 3.845569113822765e-06, "loss": 0.0085, "step": 15395 }, { "epoch": 4.78, "learning_rate": 3.8451690338067615e-06, "loss": 0.0136, "step": 15396 }, { "epoch": 4.78, "learning_rate": 3.844768953790758e-06, "loss": 0.0495, "step": 15397 }, { "epoch": 4.78, "learning_rate": 3.844368873774756e-06, "loss": 0.0322, "step": 15398 }, { "epoch": 4.78, "learning_rate": 3.843968793758752e-06, "loss": 0.0099, "step": 15399 }, { "epoch": 4.78, "learning_rate": 3.843568713742749e-06, "loss": 0.0197, "step": 15400 }, { "epoch": 4.78, "learning_rate": 3.843168633726746e-06, "loss": 0.0132, "step": 15401 }, { "epoch": 4.78, "learning_rate": 3.842768553710742e-06, "loss": 0.0072, "step": 15402 }, { "epoch": 4.78, "learning_rate": 3.842368473694739e-06, "loss": 0.0266, "step": 15403 }, { "epoch": 4.78, "learning_rate": 3.841968393678736e-06, "loss": 0.0049, "step": 15404 }, { "epoch": 4.78, "learning_rate": 3.841568313662733e-06, "loss": 0.009, "step": 15405 }, { "epoch": 4.78, "learning_rate": 3.84116823364673e-06, "loss": 0.0307, "step": 15406 }, { "epoch": 4.78, "learning_rate": 3.840768153630726e-06, "loss": 0.1107, "step": 15407 }, { "epoch": 4.78, "learning_rate": 3.840368073614723e-06, "loss": 0.0711, "step": 15408 }, { "epoch": 4.78, "learning_rate": 3.83996799359872e-06, "loss": 0.0272, "step": 15409 }, { "epoch": 4.78, "learning_rate": 3.839567913582717e-06, "loss": 0.0244, "step": 15410 }, { "epoch": 4.78, "learning_rate": 3.8391678335667134e-06, "loss": 0.0297, "step": 15411 }, { "epoch": 4.78, "learning_rate": 3.838767753550711e-06, "loss": 0.0191, "step": 15412 }, { "epoch": 4.78, "learning_rate": 3.838367673534708e-06, "loss": 0.0149, "step": 15413 }, { "epoch": 4.78, "learning_rate": 3.8379675935187035e-06, "loss": 0.028, "step": 15414 }, { "epoch": 4.78, "learning_rate": 3.837567513502701e-06, "loss": 0.0159, "step": 15415 }, { "epoch": 4.78, "learning_rate": 3.837167433486698e-06, "loss": 0.0271, "step": 15416 }, { "epoch": 4.79, "learning_rate": 3.836767353470694e-06, "loss": 0.0323, "step": 15417 }, { "epoch": 4.79, "learning_rate": 3.836367273454691e-06, "loss": 0.0088, "step": 15418 }, { "epoch": 4.79, "learning_rate": 3.835967193438689e-06, "loss": 0.0219, "step": 15419 }, { "epoch": 4.79, "learning_rate": 3.8355671134226845e-06, "loss": 0.0101, "step": 15420 }, { "epoch": 4.79, "learning_rate": 3.835167033406681e-06, "loss": 0.0153, "step": 15421 }, { "epoch": 4.79, "learning_rate": 3.834766953390679e-06, "loss": 0.0826, "step": 15422 }, { "epoch": 4.79, "learning_rate": 3.834366873374675e-06, "loss": 0.0106, "step": 15423 }, { "epoch": 4.79, "learning_rate": 3.833966793358672e-06, "loss": 0.1014, "step": 15424 }, { "epoch": 4.79, "learning_rate": 3.833566713342669e-06, "loss": 0.0274, "step": 15425 }, { "epoch": 4.79, "learning_rate": 3.8331666333266654e-06, "loss": 0.0365, "step": 15426 }, { "epoch": 4.79, "learning_rate": 3.832766553310662e-06, "loss": 0.0824, "step": 15427 }, { "epoch": 4.79, "learning_rate": 3.832366473294659e-06, "loss": 0.0113, "step": 15428 }, { "epoch": 4.79, "learning_rate": 3.831966393278656e-06, "loss": 0.0346, "step": 15429 }, { "epoch": 4.79, "learning_rate": 3.831566313262653e-06, "loss": 0.0258, "step": 15430 }, { "epoch": 4.79, "learning_rate": 3.83116623324665e-06, "loss": 0.0279, "step": 15431 }, { "epoch": 4.79, "learning_rate": 3.830766153230646e-06, "loss": 0.0202, "step": 15432 }, { "epoch": 4.79, "learning_rate": 3.830366073214643e-06, "loss": 0.0199, "step": 15433 }, { "epoch": 4.79, "learning_rate": 3.82996599319864e-06, "loss": 0.0104, "step": 15434 }, { "epoch": 4.79, "learning_rate": 3.8295659131826365e-06, "loss": 0.0097, "step": 15435 }, { "epoch": 4.79, "learning_rate": 3.829165833166634e-06, "loss": 0.0178, "step": 15436 }, { "epoch": 4.79, "learning_rate": 3.828765753150631e-06, "loss": 0.0409, "step": 15437 }, { "epoch": 4.79, "learning_rate": 3.828365673134627e-06, "loss": 0.0371, "step": 15438 }, { "epoch": 4.79, "learning_rate": 3.827965593118624e-06, "loss": 0.0954, "step": 15439 }, { "epoch": 4.79, "learning_rate": 3.827565513102621e-06, "loss": 0.016, "step": 15440 }, { "epoch": 4.79, "learning_rate": 3.827165433086617e-06, "loss": 0.0188, "step": 15441 }, { "epoch": 4.79, "learning_rate": 3.826765353070614e-06, "loss": 0.0249, "step": 15442 }, { "epoch": 4.79, "learning_rate": 3.826365273054612e-06, "loss": 0.0222, "step": 15443 }, { "epoch": 4.79, "learning_rate": 3.825965193038608e-06, "loss": 0.0127, "step": 15444 }, { "epoch": 4.79, "learning_rate": 3.825565113022604e-06, "loss": 0.1142, "step": 15445 }, { "epoch": 4.79, "learning_rate": 3.825165033006602e-06, "loss": 0.04, "step": 15446 }, { "epoch": 4.79, "learning_rate": 3.824764952990598e-06, "loss": 0.007, "step": 15447 }, { "epoch": 4.79, "learning_rate": 3.824364872974595e-06, "loss": 0.0321, "step": 15448 }, { "epoch": 4.8, "learning_rate": 3.823964792958592e-06, "loss": 0.0159, "step": 15449 }, { "epoch": 4.8, "learning_rate": 3.823564712942589e-06, "loss": 0.0207, "step": 15450 }, { "epoch": 4.8, "learning_rate": 3.823164632926586e-06, "loss": 0.0219, "step": 15451 }, { "epoch": 4.8, "learning_rate": 3.822764552910583e-06, "loss": 0.0099, "step": 15452 }, { "epoch": 4.8, "learning_rate": 3.822364472894579e-06, "loss": 0.0185, "step": 15453 }, { "epoch": 4.8, "learning_rate": 3.821964392878576e-06, "loss": 0.0179, "step": 15454 }, { "epoch": 4.8, "learning_rate": 3.821564312862573e-06, "loss": 0.079, "step": 15455 }, { "epoch": 4.8, "learning_rate": 3.821164232846569e-06, "loss": 0.0091, "step": 15456 }, { "epoch": 4.8, "learning_rate": 3.820764152830567e-06, "loss": 0.037, "step": 15457 }, { "epoch": 4.8, "learning_rate": 3.820364072814563e-06, "loss": 0.0245, "step": 15458 }, { "epoch": 4.8, "learning_rate": 3.81996399279856e-06, "loss": 0.0163, "step": 15459 }, { "epoch": 4.8, "learning_rate": 3.819563912782557e-06, "loss": 0.0334, "step": 15460 }, { "epoch": 4.8, "learning_rate": 3.819163832766554e-06, "loss": 0.0166, "step": 15461 }, { "epoch": 4.8, "learning_rate": 3.81876375275055e-06, "loss": 0.0056, "step": 15462 }, { "epoch": 4.8, "learning_rate": 3.818363672734547e-06, "loss": 0.021, "step": 15463 }, { "epoch": 4.8, "learning_rate": 3.817963592718544e-06, "loss": 0.0247, "step": 15464 }, { "epoch": 4.8, "learning_rate": 3.8175635127025404e-06, "loss": 0.0289, "step": 15465 }, { "epoch": 4.8, "learning_rate": 3.817163432686538e-06, "loss": 0.0124, "step": 15466 }, { "epoch": 4.8, "learning_rate": 3.816763352670535e-06, "loss": 0.011, "step": 15467 }, { "epoch": 4.8, "learning_rate": 3.816363272654531e-06, "loss": 0.0305, "step": 15468 }, { "epoch": 4.8, "learning_rate": 3.815963192638528e-06, "loss": 0.071, "step": 15469 }, { "epoch": 4.8, "learning_rate": 3.815563112622525e-06, "loss": 0.0372, "step": 15470 }, { "epoch": 4.8, "learning_rate": 3.815163032606521e-06, "loss": 0.026, "step": 15471 }, { "epoch": 4.8, "learning_rate": 3.814762952590518e-06, "loss": 0.015, "step": 15472 }, { "epoch": 4.8, "learning_rate": 3.814362872574515e-06, "loss": 0.0059, "step": 15473 }, { "epoch": 4.8, "learning_rate": 3.8139627925585123e-06, "loss": 0.0129, "step": 15474 }, { "epoch": 4.8, "learning_rate": 3.813562712542509e-06, "loss": 0.0448, "step": 15475 }, { "epoch": 4.8, "learning_rate": 3.8131626325265057e-06, "loss": 0.0501, "step": 15476 }, { "epoch": 4.8, "learning_rate": 3.8127625525105024e-06, "loss": 0.009, "step": 15477 }, { "epoch": 4.8, "learning_rate": 3.812362472494499e-06, "loss": 0.0106, "step": 15478 }, { "epoch": 4.8, "learning_rate": 3.811962392478496e-06, "loss": 0.0876, "step": 15479 }, { "epoch": 4.8, "learning_rate": 3.811562312462493e-06, "loss": 0.0047, "step": 15480 }, { "epoch": 4.8, "learning_rate": 3.81116223244649e-06, "loss": 0.0154, "step": 15481 }, { "epoch": 4.81, "learning_rate": 3.8107621524304866e-06, "loss": 0.0238, "step": 15482 }, { "epoch": 4.81, "learning_rate": 3.810362072414483e-06, "loss": 0.0105, "step": 15483 }, { "epoch": 4.81, "learning_rate": 3.80996199239848e-06, "loss": 0.0099, "step": 15484 }, { "epoch": 4.81, "learning_rate": 3.8095619123824767e-06, "loss": 0.0256, "step": 15485 }, { "epoch": 4.81, "learning_rate": 3.809161832366474e-06, "loss": 0.0323, "step": 15486 }, { "epoch": 4.81, "learning_rate": 3.8087617523504705e-06, "loss": 0.0313, "step": 15487 }, { "epoch": 4.81, "learning_rate": 3.8083616723344676e-06, "loss": 0.0125, "step": 15488 }, { "epoch": 4.81, "learning_rate": 3.807961592318464e-06, "loss": 0.0167, "step": 15489 }, { "epoch": 4.81, "learning_rate": 3.8075615123024606e-06, "loss": 0.0203, "step": 15490 }, { "epoch": 4.81, "learning_rate": 3.8071614322864577e-06, "loss": 0.1585, "step": 15491 }, { "epoch": 4.81, "learning_rate": 3.8067613522704543e-06, "loss": 0.0141, "step": 15492 }, { "epoch": 4.81, "learning_rate": 3.8063612722544515e-06, "loss": 0.016, "step": 15493 }, { "epoch": 4.81, "learning_rate": 3.805961192238448e-06, "loss": 0.0275, "step": 15494 }, { "epoch": 4.81, "learning_rate": 3.8055611122224453e-06, "loss": 0.0681, "step": 15495 }, { "epoch": 4.81, "learning_rate": 3.8051610322064415e-06, "loss": 0.0168, "step": 15496 }, { "epoch": 4.81, "learning_rate": 3.804760952190438e-06, "loss": 0.0105, "step": 15497 }, { "epoch": 4.81, "learning_rate": 3.8043608721744353e-06, "loss": 0.0113, "step": 15498 }, { "epoch": 4.81, "learning_rate": 3.803960792158432e-06, "loss": 0.012, "step": 15499 }, { "epoch": 4.81, "learning_rate": 3.803560712142429e-06, "loss": 0.0201, "step": 15500 }, { "epoch": 4.81, "learning_rate": 3.803160632126426e-06, "loss": 0.0096, "step": 15501 }, { "epoch": 4.81, "learning_rate": 3.802760552110422e-06, "loss": 0.0315, "step": 15502 }, { "epoch": 4.81, "learning_rate": 3.802360472094419e-06, "loss": 0.0245, "step": 15503 }, { "epoch": 4.81, "learning_rate": 3.801960392078416e-06, "loss": 0.0122, "step": 15504 }, { "epoch": 4.81, "learning_rate": 3.801560312062413e-06, "loss": 0.0061, "step": 15505 }, { "epoch": 4.81, "learning_rate": 3.8011602320464096e-06, "loss": 0.0983, "step": 15506 }, { "epoch": 4.81, "learning_rate": 3.8007601520304068e-06, "loss": 0.0433, "step": 15507 }, { "epoch": 4.81, "learning_rate": 3.800360072014403e-06, "loss": 0.0162, "step": 15508 }, { "epoch": 4.81, "learning_rate": 3.7999599919983997e-06, "loss": 0.0058, "step": 15509 }, { "epoch": 4.81, "learning_rate": 3.799559911982397e-06, "loss": 0.0354, "step": 15510 }, { "epoch": 4.81, "learning_rate": 3.7991598319663935e-06, "loss": 0.0207, "step": 15511 }, { "epoch": 4.81, "learning_rate": 3.7987597519503906e-06, "loss": 0.0201, "step": 15512 }, { "epoch": 4.81, "learning_rate": 3.7983596719343873e-06, "loss": 0.0057, "step": 15513 }, { "epoch": 4.82, "learning_rate": 3.7979595919183844e-06, "loss": 0.0287, "step": 15514 }, { "epoch": 4.82, "learning_rate": 3.7975595119023807e-06, "loss": 0.0143, "step": 15515 }, { "epoch": 4.82, "learning_rate": 3.7971594318863774e-06, "loss": 0.1353, "step": 15516 }, { "epoch": 4.82, "learning_rate": 3.7967593518703745e-06, "loss": 0.0111, "step": 15517 }, { "epoch": 4.82, "learning_rate": 3.796359271854371e-06, "loss": 0.1159, "step": 15518 }, { "epoch": 4.82, "learning_rate": 3.7959591918383683e-06, "loss": 0.0077, "step": 15519 }, { "epoch": 4.82, "learning_rate": 3.795559111822365e-06, "loss": 0.0124, "step": 15520 }, { "epoch": 4.82, "learning_rate": 3.7951590318063612e-06, "loss": 0.0256, "step": 15521 }, { "epoch": 4.82, "learning_rate": 3.7947589517903583e-06, "loss": 0.0299, "step": 15522 }, { "epoch": 4.82, "learning_rate": 3.794358871774355e-06, "loss": 0.0162, "step": 15523 }, { "epoch": 4.82, "learning_rate": 3.793958791758352e-06, "loss": 0.0113, "step": 15524 }, { "epoch": 4.82, "learning_rate": 3.793558711742349e-06, "loss": 0.0157, "step": 15525 }, { "epoch": 4.82, "learning_rate": 3.793158631726346e-06, "loss": 0.0172, "step": 15526 }, { "epoch": 4.82, "learning_rate": 3.792758551710342e-06, "loss": 0.0904, "step": 15527 }, { "epoch": 4.82, "learning_rate": 3.792358471694339e-06, "loss": 0.0289, "step": 15528 }, { "epoch": 4.82, "learning_rate": 3.791958391678336e-06, "loss": 0.0094, "step": 15529 }, { "epoch": 4.82, "learning_rate": 3.7915583116623327e-06, "loss": 0.0089, "step": 15530 }, { "epoch": 4.82, "learning_rate": 3.7911582316463298e-06, "loss": 0.0101, "step": 15531 }, { "epoch": 4.82, "learning_rate": 3.7907581516303265e-06, "loss": 0.0216, "step": 15532 }, { "epoch": 4.82, "learning_rate": 3.7903580716143227e-06, "loss": 0.02, "step": 15533 }, { "epoch": 4.82, "learning_rate": 3.78995799159832e-06, "loss": 0.0081, "step": 15534 }, { "epoch": 4.82, "learning_rate": 3.7895579115823165e-06, "loss": 0.0325, "step": 15535 }, { "epoch": 4.82, "learning_rate": 3.7891578315663136e-06, "loss": 0.0298, "step": 15536 }, { "epoch": 4.82, "learning_rate": 3.7887577515503103e-06, "loss": 0.0127, "step": 15537 }, { "epoch": 4.82, "learning_rate": 3.7883576715343074e-06, "loss": 0.003, "step": 15538 }, { "epoch": 4.82, "learning_rate": 3.787957591518304e-06, "loss": 0.0065, "step": 15539 }, { "epoch": 4.82, "learning_rate": 3.7875575115023004e-06, "loss": 0.0123, "step": 15540 }, { "epoch": 4.82, "learning_rate": 3.7871574314862975e-06, "loss": 0.0162, "step": 15541 }, { "epoch": 4.82, "learning_rate": 3.786757351470294e-06, "loss": 0.0133, "step": 15542 }, { "epoch": 4.82, "learning_rate": 3.7863572714542913e-06, "loss": 0.0118, "step": 15543 }, { "epoch": 4.82, "learning_rate": 3.785957191438288e-06, "loss": 0.0065, "step": 15544 }, { "epoch": 4.82, "learning_rate": 3.785557111422285e-06, "loss": 0.0176, "step": 15545 }, { "epoch": 4.83, "learning_rate": 3.7851570314062813e-06, "loss": 0.0094, "step": 15546 }, { "epoch": 4.83, "learning_rate": 3.784756951390278e-06, "loss": 0.0107, "step": 15547 }, { "epoch": 4.83, "learning_rate": 3.784356871374275e-06, "loss": 0.0259, "step": 15548 }, { "epoch": 4.83, "learning_rate": 3.783956791358272e-06, "loss": 0.021, "step": 15549 }, { "epoch": 4.83, "learning_rate": 3.783556711342269e-06, "loss": 0.0199, "step": 15550 }, { "epoch": 4.83, "learning_rate": 3.7831566313262656e-06, "loss": 0.0108, "step": 15551 }, { "epoch": 4.83, "learning_rate": 3.7827565513102623e-06, "loss": 0.0302, "step": 15552 }, { "epoch": 4.83, "learning_rate": 3.782356471294259e-06, "loss": 0.004, "step": 15553 }, { "epoch": 4.83, "learning_rate": 3.7819563912782557e-06, "loss": 0.0136, "step": 15554 }, { "epoch": 4.83, "learning_rate": 3.7815563112622528e-06, "loss": 0.0438, "step": 15555 }, { "epoch": 4.83, "learning_rate": 3.7811562312462495e-06, "loss": 0.0054, "step": 15556 }, { "epoch": 4.83, "learning_rate": 3.7807561512302466e-06, "loss": 0.0134, "step": 15557 }, { "epoch": 4.83, "learning_rate": 3.7803560712142433e-06, "loss": 0.0129, "step": 15558 }, { "epoch": 4.83, "learning_rate": 3.77995599119824e-06, "loss": 0.0157, "step": 15559 }, { "epoch": 4.83, "learning_rate": 3.7795559111822366e-06, "loss": 0.0193, "step": 15560 }, { "epoch": 4.83, "learning_rate": 3.7791558311662333e-06, "loss": 0.0284, "step": 15561 }, { "epoch": 4.83, "learning_rate": 3.7787557511502304e-06, "loss": 0.0902, "step": 15562 }, { "epoch": 4.83, "learning_rate": 3.778355671134227e-06, "loss": 0.0068, "step": 15563 }, { "epoch": 4.83, "learning_rate": 3.7779555911182242e-06, "loss": 0.0133, "step": 15564 }, { "epoch": 4.83, "learning_rate": 3.7775555111022205e-06, "loss": 0.0232, "step": 15565 }, { "epoch": 4.83, "learning_rate": 3.7771554310862176e-06, "loss": 0.0727, "step": 15566 }, { "epoch": 4.83, "learning_rate": 3.7767553510702143e-06, "loss": 0.0406, "step": 15567 }, { "epoch": 4.83, "learning_rate": 3.776355271054211e-06, "loss": 0.0222, "step": 15568 }, { "epoch": 4.83, "learning_rate": 3.775955191038208e-06, "loss": 0.0073, "step": 15569 }, { "epoch": 4.83, "learning_rate": 3.7755551110222048e-06, "loss": 0.0214, "step": 15570 }, { "epoch": 4.83, "learning_rate": 3.7751550310062015e-06, "loss": 0.0149, "step": 15571 }, { "epoch": 4.83, "learning_rate": 3.774754950990198e-06, "loss": 0.0338, "step": 15572 }, { "epoch": 4.83, "learning_rate": 3.7743548709741953e-06, "loss": 0.0121, "step": 15573 }, { "epoch": 4.83, "learning_rate": 3.773954790958192e-06, "loss": 0.0193, "step": 15574 }, { "epoch": 4.83, "learning_rate": 3.7735547109421886e-06, "loss": 0.0203, "step": 15575 }, { "epoch": 4.83, "learning_rate": 3.7731546309261857e-06, "loss": 0.021, "step": 15576 }, { "epoch": 4.83, "learning_rate": 3.7727545509101824e-06, "loss": 0.0155, "step": 15577 }, { "epoch": 4.84, "learning_rate": 3.772354470894179e-06, "loss": 0.0036, "step": 15578 }, { "epoch": 4.84, "learning_rate": 3.771954390878176e-06, "loss": 0.0134, "step": 15579 }, { "epoch": 4.84, "learning_rate": 3.771554310862173e-06, "loss": 0.0432, "step": 15580 }, { "epoch": 4.84, "learning_rate": 3.7711542308461696e-06, "loss": 0.012, "step": 15581 }, { "epoch": 4.84, "learning_rate": 3.7707541508301667e-06, "loss": 0.0154, "step": 15582 }, { "epoch": 4.84, "learning_rate": 3.7703540708141634e-06, "loss": 0.0184, "step": 15583 }, { "epoch": 4.84, "learning_rate": 3.7699539907981596e-06, "loss": 0.0314, "step": 15584 }, { "epoch": 4.84, "learning_rate": 3.7695539107821568e-06, "loss": 0.0364, "step": 15585 }, { "epoch": 4.84, "learning_rate": 3.7691538307661534e-06, "loss": 0.0174, "step": 15586 }, { "epoch": 4.84, "learning_rate": 3.7687537507501506e-06, "loss": 0.0192, "step": 15587 }, { "epoch": 4.84, "learning_rate": 3.7683536707341472e-06, "loss": 0.0117, "step": 15588 }, { "epoch": 4.84, "learning_rate": 3.7679535907181444e-06, "loss": 0.015, "step": 15589 }, { "epoch": 4.84, "learning_rate": 3.7675535107021406e-06, "loss": 0.0175, "step": 15590 }, { "epoch": 4.84, "learning_rate": 3.7671534306861373e-06, "loss": 0.0108, "step": 15591 }, { "epoch": 4.84, "learning_rate": 3.7667533506701344e-06, "loss": 0.0156, "step": 15592 }, { "epoch": 4.84, "learning_rate": 3.766353270654131e-06, "loss": 0.0247, "step": 15593 }, { "epoch": 4.84, "learning_rate": 3.765953190638128e-06, "loss": 0.0182, "step": 15594 }, { "epoch": 4.84, "learning_rate": 3.765553110622125e-06, "loss": 0.0338, "step": 15595 }, { "epoch": 4.84, "learning_rate": 3.765153030606122e-06, "loss": 0.0251, "step": 15596 }, { "epoch": 4.84, "learning_rate": 3.7647529505901183e-06, "loss": 0.01, "step": 15597 }, { "epoch": 4.84, "learning_rate": 3.764352870574115e-06, "loss": 0.0056, "step": 15598 }, { "epoch": 4.84, "learning_rate": 3.763952790558112e-06, "loss": 0.0045, "step": 15599 }, { "epoch": 4.84, "learning_rate": 3.7635527105421087e-06, "loss": 0.0183, "step": 15600 }, { "epoch": 4.84, "learning_rate": 3.763152630526106e-06, "loss": 0.0263, "step": 15601 }, { "epoch": 4.84, "learning_rate": 3.7627525505101025e-06, "loss": 0.0292, "step": 15602 }, { "epoch": 4.84, "learning_rate": 3.762352470494099e-06, "loss": 0.0639, "step": 15603 }, { "epoch": 4.84, "learning_rate": 3.761952390478096e-06, "loss": 0.0159, "step": 15604 }, { "epoch": 4.84, "learning_rate": 3.7615523104620926e-06, "loss": 0.0092, "step": 15605 }, { "epoch": 4.84, "learning_rate": 3.7611522304460897e-06, "loss": 0.0149, "step": 15606 }, { "epoch": 4.84, "learning_rate": 3.7607521504300864e-06, "loss": 0.0165, "step": 15607 }, { "epoch": 4.84, "learning_rate": 3.7603520704140835e-06, "loss": 0.0146, "step": 15608 }, { "epoch": 4.84, "learning_rate": 3.7599519903980798e-06, "loss": 0.0283, "step": 15609 }, { "epoch": 4.85, "learning_rate": 3.7595519103820765e-06, "loss": 0.0236, "step": 15610 }, { "epoch": 4.85, "learning_rate": 3.7591518303660736e-06, "loss": 0.0291, "step": 15611 }, { "epoch": 4.85, "learning_rate": 3.7587517503500703e-06, "loss": 0.0084, "step": 15612 }, { "epoch": 4.85, "learning_rate": 3.7583516703340674e-06, "loss": 0.0276, "step": 15613 }, { "epoch": 4.85, "learning_rate": 3.757951590318064e-06, "loss": 0.0254, "step": 15614 }, { "epoch": 4.85, "learning_rate": 3.7575515103020603e-06, "loss": 0.0493, "step": 15615 }, { "epoch": 4.85, "learning_rate": 3.7571514302860574e-06, "loss": 0.0122, "step": 15616 }, { "epoch": 4.85, "learning_rate": 3.756751350270054e-06, "loss": 0.019, "step": 15617 }, { "epoch": 4.85, "learning_rate": 3.7563512702540512e-06, "loss": 0.0085, "step": 15618 }, { "epoch": 4.85, "learning_rate": 3.755951190238048e-06, "loss": 0.016, "step": 15619 }, { "epoch": 4.85, "learning_rate": 3.755551110222045e-06, "loss": 0.0148, "step": 15620 }, { "epoch": 4.85, "learning_rate": 3.7551510302060417e-06, "loss": 0.0269, "step": 15621 }, { "epoch": 4.85, "learning_rate": 3.754750950190038e-06, "loss": 0.018, "step": 15622 }, { "epoch": 4.85, "learning_rate": 3.754350870174035e-06, "loss": 0.0302, "step": 15623 }, { "epoch": 4.85, "learning_rate": 3.7539507901580318e-06, "loss": 0.0167, "step": 15624 }, { "epoch": 4.85, "learning_rate": 3.753550710142029e-06, "loss": 0.0133, "step": 15625 }, { "epoch": 4.85, "learning_rate": 3.7531506301260256e-06, "loss": 0.1365, "step": 15626 }, { "epoch": 4.85, "learning_rate": 3.7527505501100227e-06, "loss": 0.0245, "step": 15627 }, { "epoch": 4.85, "learning_rate": 3.752350470094019e-06, "loss": 0.0241, "step": 15628 }, { "epoch": 4.85, "learning_rate": 3.7519503900780156e-06, "loss": 0.0263, "step": 15629 }, { "epoch": 4.85, "learning_rate": 3.7515503100620127e-06, "loss": 0.0092, "step": 15630 }, { "epoch": 4.85, "learning_rate": 3.7511502300460094e-06, "loss": 0.0417, "step": 15631 }, { "epoch": 4.85, "learning_rate": 3.7507501500300065e-06, "loss": 0.0124, "step": 15632 }, { "epoch": 4.85, "learning_rate": 3.750350070014003e-06, "loss": 0.0323, "step": 15633 }, { "epoch": 4.85, "learning_rate": 3.7499499899979995e-06, "loss": 0.0474, "step": 15634 }, { "epoch": 4.85, "learning_rate": 3.7495499099819966e-06, "loss": 0.009, "step": 15635 }, { "epoch": 4.85, "learning_rate": 3.7491498299659933e-06, "loss": 0.035, "step": 15636 }, { "epoch": 4.85, "learning_rate": 3.7487497499499904e-06, "loss": 0.0313, "step": 15637 }, { "epoch": 4.85, "learning_rate": 3.748349669933987e-06, "loss": 0.0184, "step": 15638 }, { "epoch": 4.85, "learning_rate": 3.747949589917984e-06, "loss": 0.0204, "step": 15639 }, { "epoch": 4.85, "learning_rate": 3.747549509901981e-06, "loss": 0.0243, "step": 15640 }, { "epoch": 4.85, "learning_rate": 3.747149429885977e-06, "loss": 0.0205, "step": 15641 }, { "epoch": 4.85, "learning_rate": 3.7467493498699742e-06, "loss": 0.043, "step": 15642 }, { "epoch": 4.86, "learning_rate": 3.746349269853971e-06, "loss": 0.0032, "step": 15643 }, { "epoch": 4.86, "learning_rate": 3.745949189837968e-06, "loss": 0.0139, "step": 15644 }, { "epoch": 4.86, "learning_rate": 3.7455491098219647e-06, "loss": 0.0283, "step": 15645 }, { "epoch": 4.86, "learning_rate": 3.745149029805962e-06, "loss": 0.0177, "step": 15646 }, { "epoch": 4.86, "learning_rate": 3.744748949789958e-06, "loss": 0.014, "step": 15647 }, { "epoch": 4.86, "learning_rate": 3.7443488697739548e-06, "loss": 0.0099, "step": 15648 }, { "epoch": 4.86, "learning_rate": 3.743948789757952e-06, "loss": 0.0091, "step": 15649 }, { "epoch": 4.86, "learning_rate": 3.7435487097419486e-06, "loss": 0.0353, "step": 15650 }, { "epoch": 4.86, "learning_rate": 3.7431486297259457e-06, "loss": 0.028, "step": 15651 }, { "epoch": 4.86, "learning_rate": 3.7427485497099424e-06, "loss": 0.0376, "step": 15652 }, { "epoch": 4.86, "learning_rate": 3.7423484696939386e-06, "loss": 0.0307, "step": 15653 }, { "epoch": 4.86, "learning_rate": 3.7419483896779357e-06, "loss": 0.0153, "step": 15654 }, { "epoch": 4.86, "learning_rate": 3.7415483096619324e-06, "loss": 0.0175, "step": 15655 }, { "epoch": 4.86, "learning_rate": 3.7411482296459295e-06, "loss": 0.0114, "step": 15656 }, { "epoch": 4.86, "learning_rate": 3.7407481496299262e-06, "loss": 0.0114, "step": 15657 }, { "epoch": 4.86, "learning_rate": 3.7403480696139233e-06, "loss": 0.017, "step": 15658 }, { "epoch": 4.86, "learning_rate": 3.73994798959792e-06, "loss": 0.0476, "step": 15659 }, { "epoch": 4.86, "learning_rate": 3.7395479095819167e-06, "loss": 0.0339, "step": 15660 }, { "epoch": 4.86, "learning_rate": 3.7391478295659134e-06, "loss": 0.0156, "step": 15661 }, { "epoch": 4.86, "learning_rate": 3.73874774954991e-06, "loss": 0.0173, "step": 15662 }, { "epoch": 4.86, "learning_rate": 3.738347669533907e-06, "loss": 0.0124, "step": 15663 }, { "epoch": 4.86, "learning_rate": 3.737947589517904e-06, "loss": 0.0095, "step": 15664 }, { "epoch": 4.86, "learning_rate": 3.737547509501901e-06, "loss": 0.0056, "step": 15665 }, { "epoch": 4.86, "learning_rate": 3.7371474294858972e-06, "loss": 0.0071, "step": 15666 }, { "epoch": 4.86, "learning_rate": 3.7367473494698944e-06, "loss": 0.0261, "step": 15667 }, { "epoch": 4.86, "learning_rate": 3.736347269453891e-06, "loss": 0.0069, "step": 15668 }, { "epoch": 4.86, "learning_rate": 3.7359471894378877e-06, "loss": 0.0096, "step": 15669 }, { "epoch": 4.86, "learning_rate": 3.735547109421885e-06, "loss": 0.0115, "step": 15670 }, { "epoch": 4.86, "learning_rate": 3.7351470294058815e-06, "loss": 0.0165, "step": 15671 }, { "epoch": 4.86, "learning_rate": 3.734746949389878e-06, "loss": 0.0165, "step": 15672 }, { "epoch": 4.86, "learning_rate": 3.734346869373875e-06, "loss": 0.0623, "step": 15673 }, { "epoch": 4.86, "learning_rate": 3.733946789357872e-06, "loss": 0.0382, "step": 15674 }, { "epoch": 4.87, "learning_rate": 3.7335467093418687e-06, "loss": 0.0693, "step": 15675 }, { "epoch": 4.87, "learning_rate": 3.7331466293258654e-06, "loss": 0.0034, "step": 15676 }, { "epoch": 4.87, "learning_rate": 3.7327465493098625e-06, "loss": 0.017, "step": 15677 }, { "epoch": 4.87, "learning_rate": 3.732346469293859e-06, "loss": 0.0206, "step": 15678 }, { "epoch": 4.87, "learning_rate": 3.731946389277856e-06, "loss": 0.0302, "step": 15679 }, { "epoch": 4.87, "learning_rate": 3.7315463092618525e-06, "loss": 0.0144, "step": 15680 }, { "epoch": 4.87, "learning_rate": 3.7311462292458497e-06, "loss": 0.0194, "step": 15681 }, { "epoch": 4.87, "learning_rate": 3.7307461492298463e-06, "loss": 0.0217, "step": 15682 }, { "epoch": 4.87, "learning_rate": 3.730346069213843e-06, "loss": 0.0695, "step": 15683 }, { "epoch": 4.87, "learning_rate": 3.72994598919784e-06, "loss": 0.0123, "step": 15684 }, { "epoch": 4.87, "learning_rate": 3.7295459091818364e-06, "loss": 0.0131, "step": 15685 }, { "epoch": 4.87, "learning_rate": 3.7291458291658335e-06, "loss": 0.0244, "step": 15686 }, { "epoch": 4.87, "learning_rate": 3.72874574914983e-06, "loss": 0.0959, "step": 15687 }, { "epoch": 4.87, "learning_rate": 3.7283456691338273e-06, "loss": 0.0378, "step": 15688 }, { "epoch": 4.87, "learning_rate": 3.727945589117824e-06, "loss": 0.0155, "step": 15689 }, { "epoch": 4.87, "learning_rate": 3.727545509101821e-06, "loss": 0.0163, "step": 15690 }, { "epoch": 4.87, "learning_rate": 3.7271454290858174e-06, "loss": 0.0267, "step": 15691 }, { "epoch": 4.87, "learning_rate": 3.726745349069814e-06, "loss": 0.0433, "step": 15692 }, { "epoch": 4.87, "learning_rate": 3.726345269053811e-06, "loss": 0.0677, "step": 15693 }, { "epoch": 4.87, "learning_rate": 3.725945189037808e-06, "loss": 0.0246, "step": 15694 }, { "epoch": 4.87, "learning_rate": 3.725545109021805e-06, "loss": 0.0184, "step": 15695 }, { "epoch": 4.87, "learning_rate": 3.7251450290058016e-06, "loss": 0.0171, "step": 15696 }, { "epoch": 4.87, "learning_rate": 3.724744948989798e-06, "loss": 0.0379, "step": 15697 }, { "epoch": 4.87, "learning_rate": 3.724344868973795e-06, "loss": 0.0128, "step": 15698 }, { "epoch": 4.87, "learning_rate": 3.7239447889577917e-06, "loss": 0.0133, "step": 15699 }, { "epoch": 4.87, "learning_rate": 3.723544708941789e-06, "loss": 0.0192, "step": 15700 }, { "epoch": 4.87, "learning_rate": 3.7231446289257855e-06, "loss": 0.0164, "step": 15701 }, { "epoch": 4.87, "learning_rate": 3.7227445489097826e-06, "loss": 0.0113, "step": 15702 }, { "epoch": 4.87, "learning_rate": 3.7223444688937793e-06, "loss": 0.0343, "step": 15703 }, { "epoch": 4.87, "learning_rate": 3.7219443888777756e-06, "loss": 0.0091, "step": 15704 }, { "epoch": 4.87, "learning_rate": 3.7215443088617727e-06, "loss": 0.0925, "step": 15705 }, { "epoch": 4.87, "learning_rate": 3.7211442288457694e-06, "loss": 0.0128, "step": 15706 }, { "epoch": 4.88, "learning_rate": 3.7207441488297665e-06, "loss": 0.0353, "step": 15707 }, { "epoch": 4.88, "learning_rate": 3.720344068813763e-06, "loss": 0.0398, "step": 15708 }, { "epoch": 4.88, "learning_rate": 3.7199439887977603e-06, "loss": 0.0046, "step": 15709 }, { "epoch": 4.88, "learning_rate": 3.7195439087817565e-06, "loss": 0.0088, "step": 15710 }, { "epoch": 4.88, "learning_rate": 3.719143828765753e-06, "loss": 0.0203, "step": 15711 }, { "epoch": 4.88, "learning_rate": 3.7187437487497503e-06, "loss": 0.0165, "step": 15712 }, { "epoch": 4.88, "learning_rate": 3.718343668733747e-06, "loss": 0.048, "step": 15713 }, { "epoch": 4.88, "learning_rate": 3.717943588717744e-06, "loss": 0.0062, "step": 15714 }, { "epoch": 4.88, "learning_rate": 3.717543508701741e-06, "loss": 0.0173, "step": 15715 }, { "epoch": 4.88, "learning_rate": 3.717143428685737e-06, "loss": 0.0295, "step": 15716 }, { "epoch": 4.88, "learning_rate": 3.716743348669734e-06, "loss": 0.039, "step": 15717 }, { "epoch": 4.88, "learning_rate": 3.716343268653731e-06, "loss": 0.0097, "step": 15718 }, { "epoch": 4.88, "learning_rate": 3.715943188637728e-06, "loss": 0.0116, "step": 15719 }, { "epoch": 4.88, "learning_rate": 3.7155431086217247e-06, "loss": 0.0135, "step": 15720 }, { "epoch": 4.88, "learning_rate": 3.7151430286057218e-06, "loss": 0.0168, "step": 15721 }, { "epoch": 4.88, "learning_rate": 3.7147429485897184e-06, "loss": 0.0328, "step": 15722 }, { "epoch": 4.88, "learning_rate": 3.7143428685737147e-06, "loss": 0.0274, "step": 15723 }, { "epoch": 4.88, "learning_rate": 3.713942788557712e-06, "loss": 0.016, "step": 15724 }, { "epoch": 4.88, "learning_rate": 3.7135427085417085e-06, "loss": 0.0071, "step": 15725 }, { "epoch": 4.88, "learning_rate": 3.7131426285257056e-06, "loss": 0.0103, "step": 15726 }, { "epoch": 4.88, "learning_rate": 3.7127425485097023e-06, "loss": 0.0613, "step": 15727 }, { "epoch": 4.88, "learning_rate": 3.7123424684936994e-06, "loss": 0.004, "step": 15728 }, { "epoch": 4.88, "learning_rate": 3.7119423884776957e-06, "loss": 0.0178, "step": 15729 }, { "epoch": 4.88, "learning_rate": 3.7115423084616924e-06, "loss": 0.0148, "step": 15730 }, { "epoch": 4.88, "learning_rate": 3.7111422284456895e-06, "loss": 0.0086, "step": 15731 }, { "epoch": 4.88, "learning_rate": 3.710742148429686e-06, "loss": 0.027, "step": 15732 }, { "epoch": 4.88, "learning_rate": 3.7103420684136833e-06, "loss": 0.0112, "step": 15733 }, { "epoch": 4.88, "learning_rate": 3.70994198839768e-06, "loss": 0.0335, "step": 15734 }, { "epoch": 4.88, "learning_rate": 3.7095419083816762e-06, "loss": 0.0069, "step": 15735 }, { "epoch": 4.88, "learning_rate": 3.7091418283656733e-06, "loss": 0.016, "step": 15736 }, { "epoch": 4.88, "learning_rate": 3.70874174834967e-06, "loss": 0.0336, "step": 15737 }, { "epoch": 4.88, "learning_rate": 3.708341668333667e-06, "loss": 0.0132, "step": 15738 }, { "epoch": 4.89, "learning_rate": 3.707941588317664e-06, "loss": 0.1004, "step": 15739 }, { "epoch": 4.89, "learning_rate": 3.707541508301661e-06, "loss": 0.0143, "step": 15740 }, { "epoch": 4.89, "learning_rate": 3.7071414282856576e-06, "loss": 0.014, "step": 15741 }, { "epoch": 4.89, "learning_rate": 3.706741348269654e-06, "loss": 0.0123, "step": 15742 }, { "epoch": 4.89, "learning_rate": 3.706341268253651e-06, "loss": 0.0106, "step": 15743 }, { "epoch": 4.89, "learning_rate": 3.7059411882376477e-06, "loss": 0.0184, "step": 15744 }, { "epoch": 4.89, "learning_rate": 3.7055411082216448e-06, "loss": 0.0277, "step": 15745 }, { "epoch": 4.89, "learning_rate": 3.7051410282056415e-06, "loss": 0.0321, "step": 15746 }, { "epoch": 4.89, "learning_rate": 3.7047409481896386e-06, "loss": 0.0049, "step": 15747 }, { "epoch": 4.89, "learning_rate": 3.704340868173635e-06, "loss": 0.0072, "step": 15748 }, { "epoch": 4.89, "learning_rate": 3.7039407881576315e-06, "loss": 0.0162, "step": 15749 }, { "epoch": 4.89, "learning_rate": 3.7035407081416286e-06, "loss": 0.0141, "step": 15750 }, { "epoch": 4.89, "learning_rate": 3.7031406281256253e-06, "loss": 0.0196, "step": 15751 }, { "epoch": 4.89, "learning_rate": 3.7027405481096224e-06, "loss": 0.0281, "step": 15752 }, { "epoch": 4.89, "learning_rate": 3.702340468093619e-06, "loss": 0.02, "step": 15753 }, { "epoch": 4.89, "learning_rate": 3.7019403880776154e-06, "loss": 0.0302, "step": 15754 }, { "epoch": 4.89, "learning_rate": 3.7015403080616125e-06, "loss": 0.0062, "step": 15755 }, { "epoch": 4.89, "learning_rate": 3.701140228045609e-06, "loss": 0.0186, "step": 15756 }, { "epoch": 4.89, "learning_rate": 3.7007401480296063e-06, "loss": 0.0111, "step": 15757 }, { "epoch": 4.89, "learning_rate": 3.700340068013603e-06, "loss": 0.008, "step": 15758 }, { "epoch": 4.89, "learning_rate": 3.6999399879976e-06, "loss": 0.0081, "step": 15759 }, { "epoch": 4.89, "learning_rate": 3.6995399079815963e-06, "loss": 0.0246, "step": 15760 }, { "epoch": 4.89, "learning_rate": 3.699139827965593e-06, "loss": 0.0184, "step": 15761 }, { "epoch": 4.89, "learning_rate": 3.69873974794959e-06, "loss": 0.0112, "step": 15762 }, { "epoch": 4.89, "learning_rate": 3.698339667933587e-06, "loss": 0.0264, "step": 15763 }, { "epoch": 4.89, "learning_rate": 3.697939587917584e-06, "loss": 0.0058, "step": 15764 }, { "epoch": 4.89, "learning_rate": 3.6975395079015806e-06, "loss": 0.0216, "step": 15765 }, { "epoch": 4.89, "learning_rate": 3.6971394278855777e-06, "loss": 0.0298, "step": 15766 }, { "epoch": 4.89, "learning_rate": 3.696739347869574e-06, "loss": 0.0122, "step": 15767 }, { "epoch": 4.89, "learning_rate": 3.696339267853571e-06, "loss": 0.0251, "step": 15768 }, { "epoch": 4.89, "learning_rate": 3.6959391878375678e-06, "loss": 0.0098, "step": 15769 }, { "epoch": 4.89, "learning_rate": 3.6955391078215645e-06, "loss": 0.0751, "step": 15770 }, { "epoch": 4.89, "learning_rate": 3.6951390278055616e-06, "loss": 0.0057, "step": 15771 }, { "epoch": 4.9, "learning_rate": 3.6947389477895583e-06, "loss": 0.0196, "step": 15772 }, { "epoch": 4.9, "learning_rate": 3.694338867773555e-06, "loss": 0.0057, "step": 15773 }, { "epoch": 4.9, "learning_rate": 3.6939387877575516e-06, "loss": 0.0141, "step": 15774 }, { "epoch": 4.9, "learning_rate": 3.6935387077415488e-06, "loss": 0.0142, "step": 15775 }, { "epoch": 4.9, "learning_rate": 3.6931386277255454e-06, "loss": 0.0216, "step": 15776 }, { "epoch": 4.9, "learning_rate": 3.692738547709542e-06, "loss": 0.0359, "step": 15777 }, { "epoch": 4.9, "learning_rate": 3.6923384676935392e-06, "loss": 0.0267, "step": 15778 }, { "epoch": 4.9, "learning_rate": 3.6919383876775355e-06, "loss": 0.0302, "step": 15779 }, { "epoch": 4.9, "learning_rate": 3.6915383076615326e-06, "loss": 0.0245, "step": 15780 }, { "epoch": 4.9, "learning_rate": 3.6911382276455293e-06, "loss": 0.0197, "step": 15781 }, { "epoch": 4.9, "learning_rate": 3.6907381476295264e-06, "loss": 0.0362, "step": 15782 }, { "epoch": 4.9, "learning_rate": 3.690338067613523e-06, "loss": 0.015, "step": 15783 }, { "epoch": 4.9, "learning_rate": 3.6899379875975198e-06, "loss": 0.027, "step": 15784 }, { "epoch": 4.9, "learning_rate": 3.689537907581517e-06, "loss": 0.0133, "step": 15785 }, { "epoch": 4.9, "learning_rate": 3.689137827565513e-06, "loss": 0.0108, "step": 15786 }, { "epoch": 4.9, "learning_rate": 3.6887377475495103e-06, "loss": 0.0151, "step": 15787 }, { "epoch": 4.9, "learning_rate": 3.688337667533507e-06, "loss": 0.0225, "step": 15788 }, { "epoch": 4.9, "learning_rate": 3.687937587517504e-06, "loss": 0.0119, "step": 15789 }, { "epoch": 4.9, "learning_rate": 3.6875375075015007e-06, "loss": 0.033, "step": 15790 }, { "epoch": 4.9, "learning_rate": 3.6871374274854974e-06, "loss": 0.0211, "step": 15791 }, { "epoch": 4.9, "learning_rate": 3.686737347469494e-06, "loss": 0.0121, "step": 15792 }, { "epoch": 4.9, "learning_rate": 3.686337267453491e-06, "loss": 0.023, "step": 15793 }, { "epoch": 4.9, "learning_rate": 3.685937187437488e-06, "loss": 0.0115, "step": 15794 }, { "epoch": 4.9, "learning_rate": 3.6855371074214846e-06, "loss": 0.007, "step": 15795 }, { "epoch": 4.9, "learning_rate": 3.6851370274054817e-06, "loss": 0.0065, "step": 15796 }, { "epoch": 4.9, "learning_rate": 3.6847369473894784e-06, "loss": 0.008, "step": 15797 }, { "epoch": 4.9, "learning_rate": 3.6843368673734747e-06, "loss": 0.0661, "step": 15798 }, { "epoch": 4.9, "learning_rate": 3.6839367873574718e-06, "loss": 0.0215, "step": 15799 }, { "epoch": 4.9, "learning_rate": 3.6835367073414684e-06, "loss": 0.0236, "step": 15800 }, { "epoch": 4.9, "learning_rate": 3.6831366273254656e-06, "loss": 0.0671, "step": 15801 }, { "epoch": 4.9, "learning_rate": 3.6827365473094622e-06, "loss": 0.0182, "step": 15802 }, { "epoch": 4.9, "learning_rate": 3.6823364672934594e-06, "loss": 0.013, "step": 15803 }, { "epoch": 4.91, "learning_rate": 3.681936387277456e-06, "loss": 0.0193, "step": 15804 }, { "epoch": 4.91, "learning_rate": 3.6815363072614523e-06, "loss": 0.0323, "step": 15805 }, { "epoch": 4.91, "learning_rate": 3.6811362272454494e-06, "loss": 0.0552, "step": 15806 }, { "epoch": 4.91, "learning_rate": 3.680736147229446e-06, "loss": 0.0118, "step": 15807 }, { "epoch": 4.91, "learning_rate": 3.680336067213443e-06, "loss": 0.0041, "step": 15808 }, { "epoch": 4.91, "learning_rate": 3.67993598719744e-06, "loss": 0.0196, "step": 15809 }, { "epoch": 4.91, "learning_rate": 3.679535907181437e-06, "loss": 0.0074, "step": 15810 }, { "epoch": 4.91, "learning_rate": 3.6791358271654333e-06, "loss": 0.0051, "step": 15811 }, { "epoch": 4.91, "learning_rate": 3.67873574714943e-06, "loss": 0.0088, "step": 15812 }, { "epoch": 4.91, "learning_rate": 3.678335667133427e-06, "loss": 0.0155, "step": 15813 }, { "epoch": 4.91, "learning_rate": 3.6779355871174238e-06, "loss": 0.0061, "step": 15814 }, { "epoch": 4.91, "learning_rate": 3.677535507101421e-06, "loss": 0.0166, "step": 15815 }, { "epoch": 4.91, "learning_rate": 3.6771354270854175e-06, "loss": 0.0273, "step": 15816 }, { "epoch": 4.91, "learning_rate": 3.676735347069414e-06, "loss": 0.0087, "step": 15817 }, { "epoch": 4.91, "learning_rate": 3.676335267053411e-06, "loss": 0.0198, "step": 15818 }, { "epoch": 4.91, "learning_rate": 3.6759351870374076e-06, "loss": 0.0112, "step": 15819 }, { "epoch": 4.91, "learning_rate": 3.6755351070214047e-06, "loss": 0.0084, "step": 15820 }, { "epoch": 4.91, "learning_rate": 3.6751350270054014e-06, "loss": 0.0182, "step": 15821 }, { "epoch": 4.91, "learning_rate": 3.6747349469893985e-06, "loss": 0.026, "step": 15822 }, { "epoch": 4.91, "learning_rate": 3.674334866973395e-06, "loss": 0.011, "step": 15823 }, { "epoch": 4.91, "learning_rate": 3.6739347869573915e-06, "loss": 0.0375, "step": 15824 }, { "epoch": 4.91, "learning_rate": 3.6735347069413886e-06, "loss": 0.02, "step": 15825 }, { "epoch": 4.91, "learning_rate": 3.6731346269253853e-06, "loss": 0.0112, "step": 15826 }, { "epoch": 4.91, "learning_rate": 3.6727345469093824e-06, "loss": 0.026, "step": 15827 }, { "epoch": 4.91, "learning_rate": 3.672334466893379e-06, "loss": 0.0114, "step": 15828 }, { "epoch": 4.91, "learning_rate": 3.671934386877376e-06, "loss": 0.0134, "step": 15829 }, { "epoch": 4.91, "learning_rate": 3.6715343068613724e-06, "loss": 0.015, "step": 15830 }, { "epoch": 4.91, "learning_rate": 3.671134226845369e-06, "loss": 0.0247, "step": 15831 }, { "epoch": 4.91, "learning_rate": 3.6707341468293662e-06, "loss": 0.0184, "step": 15832 }, { "epoch": 4.91, "learning_rate": 3.670334066813363e-06, "loss": 0.0488, "step": 15833 }, { "epoch": 4.91, "learning_rate": 3.66993398679736e-06, "loss": 0.016, "step": 15834 }, { "epoch": 4.91, "learning_rate": 3.6695339067813567e-06, "loss": 0.0135, "step": 15835 }, { "epoch": 4.92, "learning_rate": 3.669133826765353e-06, "loss": 0.0336, "step": 15836 }, { "epoch": 4.92, "learning_rate": 3.66873374674935e-06, "loss": 0.0234, "step": 15837 }, { "epoch": 4.92, "learning_rate": 3.6683336667333468e-06, "loss": 0.0119, "step": 15838 }, { "epoch": 4.92, "learning_rate": 3.667933586717344e-06, "loss": 0.0134, "step": 15839 }, { "epoch": 4.92, "learning_rate": 3.6675335067013406e-06, "loss": 0.0133, "step": 15840 }, { "epoch": 4.92, "learning_rate": 3.6671334266853377e-06, "loss": 0.0129, "step": 15841 }, { "epoch": 4.92, "learning_rate": 3.666733346669334e-06, "loss": 0.024, "step": 15842 }, { "epoch": 4.92, "learning_rate": 3.6663332666533306e-06, "loss": 0.0257, "step": 15843 }, { "epoch": 4.92, "learning_rate": 3.6659331866373277e-06, "loss": 0.0208, "step": 15844 }, { "epoch": 4.92, "learning_rate": 3.6655331066213244e-06, "loss": 0.0179, "step": 15845 }, { "epoch": 4.92, "learning_rate": 3.6651330266053215e-06, "loss": 0.0126, "step": 15846 }, { "epoch": 4.92, "learning_rate": 3.664732946589318e-06, "loss": 0.0039, "step": 15847 }, { "epoch": 4.92, "learning_rate": 3.6643328665733153e-06, "loss": 0.0169, "step": 15848 }, { "epoch": 4.92, "learning_rate": 3.6639327865573116e-06, "loss": 0.0142, "step": 15849 }, { "epoch": 4.92, "learning_rate": 3.6635327065413083e-06, "loss": 0.0126, "step": 15850 }, { "epoch": 4.92, "learning_rate": 3.6631326265253054e-06, "loss": 0.012, "step": 15851 }, { "epoch": 4.92, "learning_rate": 3.662732546509302e-06, "loss": 0.0484, "step": 15852 }, { "epoch": 4.92, "learning_rate": 3.662332466493299e-06, "loss": 0.0158, "step": 15853 }, { "epoch": 4.92, "learning_rate": 3.661932386477296e-06, "loss": 0.0343, "step": 15854 }, { "epoch": 4.92, "learning_rate": 3.661532306461292e-06, "loss": 0.0142, "step": 15855 }, { "epoch": 4.92, "learning_rate": 3.6611322264452892e-06, "loss": 0.0234, "step": 15856 }, { "epoch": 4.92, "learning_rate": 3.660732146429286e-06, "loss": 0.0196, "step": 15857 }, { "epoch": 4.92, "learning_rate": 3.660332066413283e-06, "loss": 0.0074, "step": 15858 }, { "epoch": 4.92, "learning_rate": 3.6599319863972797e-06, "loss": 0.0182, "step": 15859 }, { "epoch": 4.92, "learning_rate": 3.659531906381277e-06, "loss": 0.009, "step": 15860 }, { "epoch": 4.92, "learning_rate": 3.659131826365273e-06, "loss": 0.0264, "step": 15861 }, { "epoch": 4.92, "learning_rate": 3.6587317463492698e-06, "loss": 0.0271, "step": 15862 }, { "epoch": 4.92, "learning_rate": 3.658331666333267e-06, "loss": 0.0184, "step": 15863 }, { "epoch": 4.92, "learning_rate": 3.6579315863172636e-06, "loss": 0.0075, "step": 15864 }, { "epoch": 4.92, "learning_rate": 3.6575315063012607e-06, "loss": 0.0077, "step": 15865 }, { "epoch": 4.92, "learning_rate": 3.6571314262852574e-06, "loss": 0.0358, "step": 15866 }, { "epoch": 4.92, "learning_rate": 3.6567313462692545e-06, "loss": 0.0043, "step": 15867 }, { "epoch": 4.93, "learning_rate": 3.6563312662532507e-06, "loss": 0.0191, "step": 15868 }, { "epoch": 4.93, "learning_rate": 3.6559311862372474e-06, "loss": 0.0298, "step": 15869 }, { "epoch": 4.93, "learning_rate": 3.6555311062212445e-06, "loss": 0.0717, "step": 15870 }, { "epoch": 4.93, "learning_rate": 3.6551310262052412e-06, "loss": 0.0161, "step": 15871 }, { "epoch": 4.93, "learning_rate": 3.6547309461892383e-06, "loss": 0.0903, "step": 15872 }, { "epoch": 4.93, "learning_rate": 3.654330866173235e-06, "loss": 0.0245, "step": 15873 }, { "epoch": 4.93, "learning_rate": 3.6539307861572317e-06, "loss": 0.005, "step": 15874 }, { "epoch": 4.93, "learning_rate": 3.6535307061412284e-06, "loss": 0.0436, "step": 15875 }, { "epoch": 4.93, "learning_rate": 3.653130626125225e-06, "loss": 0.0044, "step": 15876 }, { "epoch": 4.93, "learning_rate": 3.652730546109222e-06, "loss": 0.0311, "step": 15877 }, { "epoch": 4.93, "learning_rate": 3.652330466093219e-06, "loss": 0.0241, "step": 15878 }, { "epoch": 4.93, "learning_rate": 3.651930386077216e-06, "loss": 0.0092, "step": 15879 }, { "epoch": 4.93, "learning_rate": 3.6515303060612122e-06, "loss": 0.0107, "step": 15880 }, { "epoch": 4.93, "learning_rate": 3.6511302260452094e-06, "loss": 0.0194, "step": 15881 }, { "epoch": 4.93, "learning_rate": 3.650730146029206e-06, "loss": 0.0042, "step": 15882 }, { "epoch": 4.93, "learning_rate": 3.650330066013203e-06, "loss": 0.0221, "step": 15883 }, { "epoch": 4.93, "learning_rate": 3.6499299859972e-06, "loss": 0.0084, "step": 15884 }, { "epoch": 4.93, "learning_rate": 3.6495299059811965e-06, "loss": 0.0238, "step": 15885 }, { "epoch": 4.93, "learning_rate": 3.6491298259651936e-06, "loss": 0.0302, "step": 15886 }, { "epoch": 4.93, "learning_rate": 3.64872974594919e-06, "loss": 0.02, "step": 15887 }, { "epoch": 4.93, "learning_rate": 3.648329665933187e-06, "loss": 0.0114, "step": 15888 }, { "epoch": 4.93, "learning_rate": 3.6479295859171837e-06, "loss": 0.0085, "step": 15889 }, { "epoch": 4.93, "learning_rate": 3.647529505901181e-06, "loss": 0.0121, "step": 15890 }, { "epoch": 4.93, "learning_rate": 3.6471294258851775e-06, "loss": 0.0597, "step": 15891 }, { "epoch": 4.93, "learning_rate": 3.646729345869174e-06, "loss": 0.0118, "step": 15892 }, { "epoch": 4.93, "learning_rate": 3.646329265853171e-06, "loss": 0.0313, "step": 15893 }, { "epoch": 4.93, "learning_rate": 3.6459291858371675e-06, "loss": 0.0746, "step": 15894 }, { "epoch": 4.93, "learning_rate": 3.6455291058211647e-06, "loss": 0.0145, "step": 15895 }, { "epoch": 4.93, "learning_rate": 3.6451290258051613e-06, "loss": 0.0429, "step": 15896 }, { "epoch": 4.93, "learning_rate": 3.6447289457891585e-06, "loss": 0.0289, "step": 15897 }, { "epoch": 4.93, "learning_rate": 3.644328865773155e-06, "loss": 0.0988, "step": 15898 }, { "epoch": 4.93, "learning_rate": 3.6439287857571514e-06, "loss": 0.0204, "step": 15899 }, { "epoch": 4.94, "learning_rate": 3.6435287057411485e-06, "loss": 0.0137, "step": 15900 }, { "epoch": 4.94, "learning_rate": 3.643128625725145e-06, "loss": 0.0056, "step": 15901 }, { "epoch": 4.94, "learning_rate": 3.6427285457091423e-06, "loss": 0.0296, "step": 15902 }, { "epoch": 4.94, "learning_rate": 3.642328465693139e-06, "loss": 0.0102, "step": 15903 }, { "epoch": 4.94, "learning_rate": 3.641928385677136e-06, "loss": 0.0893, "step": 15904 }, { "epoch": 4.94, "learning_rate": 3.641528305661133e-06, "loss": 0.0128, "step": 15905 }, { "epoch": 4.94, "learning_rate": 3.641128225645129e-06, "loss": 0.0346, "step": 15906 }, { "epoch": 4.94, "learning_rate": 3.640728145629126e-06, "loss": 0.0079, "step": 15907 }, { "epoch": 4.94, "learning_rate": 3.640328065613123e-06, "loss": 0.0216, "step": 15908 }, { "epoch": 4.94, "learning_rate": 3.63992798559712e-06, "loss": 0.0276, "step": 15909 }, { "epoch": 4.94, "learning_rate": 3.6395279055811166e-06, "loss": 0.0222, "step": 15910 }, { "epoch": 4.94, "learning_rate": 3.6391278255651138e-06, "loss": 0.0055, "step": 15911 }, { "epoch": 4.94, "learning_rate": 3.63872774554911e-06, "loss": 0.0387, "step": 15912 }, { "epoch": 4.94, "learning_rate": 3.6383276655331067e-06, "loss": 0.0181, "step": 15913 }, { "epoch": 4.94, "learning_rate": 3.637927585517104e-06, "loss": 0.0052, "step": 15914 }, { "epoch": 4.94, "learning_rate": 3.6375275055011005e-06, "loss": 0.0221, "step": 15915 }, { "epoch": 4.94, "learning_rate": 3.6371274254850976e-06, "loss": 0.0414, "step": 15916 }, { "epoch": 4.94, "learning_rate": 3.6367273454690943e-06, "loss": 0.0258, "step": 15917 }, { "epoch": 4.94, "learning_rate": 3.6363272654530906e-06, "loss": 0.0123, "step": 15918 }, { "epoch": 4.94, "learning_rate": 3.6359271854370877e-06, "loss": 0.052, "step": 15919 }, { "epoch": 4.94, "learning_rate": 3.6355271054210844e-06, "loss": 0.0221, "step": 15920 }, { "epoch": 4.94, "learning_rate": 3.6351270254050815e-06, "loss": 0.0085, "step": 15921 }, { "epoch": 4.94, "learning_rate": 3.634726945389078e-06, "loss": 0.007, "step": 15922 }, { "epoch": 4.94, "learning_rate": 3.6343268653730753e-06, "loss": 0.0132, "step": 15923 }, { "epoch": 4.94, "learning_rate": 3.6339267853570715e-06, "loss": 0.0235, "step": 15924 }, { "epoch": 4.94, "learning_rate": 3.633526705341068e-06, "loss": 0.0142, "step": 15925 }, { "epoch": 4.94, "learning_rate": 3.6331266253250653e-06, "loss": 0.0229, "step": 15926 }, { "epoch": 4.94, "learning_rate": 3.632726545309062e-06, "loss": 0.0095, "step": 15927 }, { "epoch": 4.94, "learning_rate": 3.632326465293059e-06, "loss": 0.0132, "step": 15928 }, { "epoch": 4.94, "learning_rate": 3.631926385277056e-06, "loss": 0.0273, "step": 15929 }, { "epoch": 4.94, "learning_rate": 3.631526305261053e-06, "loss": 0.0129, "step": 15930 }, { "epoch": 4.94, "learning_rate": 3.631126225245049e-06, "loss": 0.0049, "step": 15931 }, { "epoch": 4.94, "learning_rate": 3.630726145229046e-06, "loss": 0.0236, "step": 15932 }, { "epoch": 4.95, "learning_rate": 3.630326065213043e-06, "loss": 0.0446, "step": 15933 }, { "epoch": 4.95, "learning_rate": 3.6299259851970397e-06, "loss": 0.0452, "step": 15934 }, { "epoch": 4.95, "learning_rate": 3.6295259051810368e-06, "loss": 0.0245, "step": 15935 }, { "epoch": 4.95, "learning_rate": 3.6291258251650335e-06, "loss": 0.0039, "step": 15936 }, { "epoch": 4.95, "learning_rate": 3.6287257451490297e-06, "loss": 0.0187, "step": 15937 }, { "epoch": 4.95, "learning_rate": 3.628325665133027e-06, "loss": 0.0059, "step": 15938 }, { "epoch": 4.95, "learning_rate": 3.6279255851170235e-06, "loss": 0.0091, "step": 15939 }, { "epoch": 4.95, "learning_rate": 3.6275255051010206e-06, "loss": 0.082, "step": 15940 }, { "epoch": 4.95, "learning_rate": 3.6271254250850173e-06, "loss": 0.0667, "step": 15941 }, { "epoch": 4.95, "learning_rate": 3.6267253450690144e-06, "loss": 0.0163, "step": 15942 }, { "epoch": 4.95, "learning_rate": 3.6263252650530107e-06, "loss": 0.0249, "step": 15943 }, { "epoch": 4.95, "learning_rate": 3.6259251850370074e-06, "loss": 0.0131, "step": 15944 }, { "epoch": 4.95, "learning_rate": 3.6255251050210045e-06, "loss": 0.0172, "step": 15945 }, { "epoch": 4.95, "learning_rate": 3.625125025005001e-06, "loss": 0.0145, "step": 15946 }, { "epoch": 4.95, "learning_rate": 3.6247249449889983e-06, "loss": 0.0405, "step": 15947 }, { "epoch": 4.95, "learning_rate": 3.624324864972995e-06, "loss": 0.0516, "step": 15948 }, { "epoch": 4.95, "learning_rate": 3.623924784956992e-06, "loss": 0.0125, "step": 15949 }, { "epoch": 4.95, "learning_rate": 3.6235247049409883e-06, "loss": 0.1123, "step": 15950 }, { "epoch": 4.95, "learning_rate": 3.623124624924985e-06, "loss": 0.0932, "step": 15951 }, { "epoch": 4.95, "learning_rate": 3.622724544908982e-06, "loss": 0.0185, "step": 15952 }, { "epoch": 4.95, "learning_rate": 3.622324464892979e-06, "loss": 0.0225, "step": 15953 }, { "epoch": 4.95, "learning_rate": 3.621924384876976e-06, "loss": 0.0565, "step": 15954 }, { "epoch": 4.95, "learning_rate": 3.6215243048609726e-06, "loss": 0.0193, "step": 15955 }, { "epoch": 4.95, "learning_rate": 3.621124224844969e-06, "loss": 0.0103, "step": 15956 }, { "epoch": 4.95, "learning_rate": 3.620724144828966e-06, "loss": 0.0256, "step": 15957 }, { "epoch": 4.95, "learning_rate": 3.6203240648129627e-06, "loss": 0.0277, "step": 15958 }, { "epoch": 4.95, "learning_rate": 3.6199239847969598e-06, "loss": 0.0141, "step": 15959 }, { "epoch": 4.95, "learning_rate": 3.6195239047809565e-06, "loss": 0.0103, "step": 15960 }, { "epoch": 4.95, "learning_rate": 3.6191238247649536e-06, "loss": 0.0211, "step": 15961 }, { "epoch": 4.95, "learning_rate": 3.61872374474895e-06, "loss": 0.01, "step": 15962 }, { "epoch": 4.95, "learning_rate": 3.6183236647329465e-06, "loss": 0.014, "step": 15963 }, { "epoch": 4.95, "learning_rate": 3.6179235847169436e-06, "loss": 0.0092, "step": 15964 }, { "epoch": 4.96, "learning_rate": 3.6175235047009403e-06, "loss": 0.0183, "step": 15965 }, { "epoch": 4.96, "learning_rate": 3.6171234246849374e-06, "loss": 0.0176, "step": 15966 }, { "epoch": 4.96, "learning_rate": 3.616723344668934e-06, "loss": 0.0115, "step": 15967 }, { "epoch": 4.96, "learning_rate": 3.6163232646529312e-06, "loss": 0.0162, "step": 15968 }, { "epoch": 4.96, "learning_rate": 3.6159231846369275e-06, "loss": 0.0178, "step": 15969 }, { "epoch": 4.96, "learning_rate": 3.615523104620924e-06, "loss": 0.1422, "step": 15970 }, { "epoch": 4.96, "learning_rate": 3.6151230246049213e-06, "loss": 0.0409, "step": 15971 }, { "epoch": 4.96, "learning_rate": 3.614722944588918e-06, "loss": 0.0213, "step": 15972 }, { "epoch": 4.96, "learning_rate": 3.614322864572915e-06, "loss": 0.0224, "step": 15973 }, { "epoch": 4.96, "learning_rate": 3.6139227845569118e-06, "loss": 0.0263, "step": 15974 }, { "epoch": 4.96, "learning_rate": 3.6135227045409085e-06, "loss": 0.0068, "step": 15975 }, { "epoch": 4.96, "learning_rate": 3.613122624524905e-06, "loss": 0.1158, "step": 15976 }, { "epoch": 4.96, "learning_rate": 3.612722544508902e-06, "loss": 0.0298, "step": 15977 }, { "epoch": 4.96, "learning_rate": 3.612322464492899e-06, "loss": 0.0215, "step": 15978 }, { "epoch": 4.96, "learning_rate": 3.6119223844768956e-06, "loss": 0.0128, "step": 15979 }, { "epoch": 4.96, "learning_rate": 3.6115223044608927e-06, "loss": 0.0084, "step": 15980 }, { "epoch": 4.96, "learning_rate": 3.611122224444889e-06, "loss": 0.0067, "step": 15981 }, { "epoch": 4.96, "learning_rate": 3.610722144428886e-06, "loss": 0.0152, "step": 15982 }, { "epoch": 4.96, "learning_rate": 3.610322064412883e-06, "loss": 0.0124, "step": 15983 }, { "epoch": 4.96, "learning_rate": 3.6099219843968795e-06, "loss": 0.0454, "step": 15984 }, { "epoch": 4.96, "learning_rate": 3.6095219043808766e-06, "loss": 0.0194, "step": 15985 }, { "epoch": 4.96, "learning_rate": 3.6091218243648733e-06, "loss": 0.0142, "step": 15986 }, { "epoch": 4.96, "learning_rate": 3.60872174434887e-06, "loss": 0.0137, "step": 15987 }, { "epoch": 4.96, "learning_rate": 3.6083216643328666e-06, "loss": 0.0223, "step": 15988 }, { "epoch": 4.96, "learning_rate": 3.6079215843168638e-06, "loss": 0.0068, "step": 15989 }, { "epoch": 4.96, "learning_rate": 3.6075215043008604e-06, "loss": 0.0047, "step": 15990 }, { "epoch": 4.96, "learning_rate": 3.6071214242848576e-06, "loss": 0.0356, "step": 15991 }, { "epoch": 4.96, "learning_rate": 3.6067213442688542e-06, "loss": 0.0257, "step": 15992 }, { "epoch": 4.96, "learning_rate": 3.606321264252851e-06, "loss": 0.0217, "step": 15993 }, { "epoch": 4.96, "learning_rate": 3.6059211842368476e-06, "loss": 0.0177, "step": 15994 }, { "epoch": 4.96, "learning_rate": 3.6055211042208443e-06, "loss": 0.0252, "step": 15995 }, { "epoch": 4.96, "learning_rate": 3.6051210242048414e-06, "loss": 0.0071, "step": 15996 }, { "epoch": 4.97, "learning_rate": 3.604720944188838e-06, "loss": 0.0166, "step": 15997 }, { "epoch": 4.97, "learning_rate": 3.604320864172835e-06, "loss": 0.0047, "step": 15998 }, { "epoch": 4.97, "learning_rate": 3.603920784156832e-06, "loss": 0.0126, "step": 15999 }, { "epoch": 4.97, "learning_rate": 3.603520704140828e-06, "loss": 0.0155, "step": 16000 }, { "epoch": 4.97, "learning_rate": 3.6031206241248253e-06, "loss": 0.0307, "step": 16001 }, { "epoch": 4.97, "learning_rate": 3.602720544108822e-06, "loss": 0.0079, "step": 16002 }, { "epoch": 4.97, "learning_rate": 3.602320464092819e-06, "loss": 0.0482, "step": 16003 }, { "epoch": 4.97, "learning_rate": 3.6019203840768157e-06, "loss": 0.013, "step": 16004 }, { "epoch": 4.97, "learning_rate": 3.601520304060813e-06, "loss": 0.0207, "step": 16005 }, { "epoch": 4.97, "learning_rate": 3.601120224044809e-06, "loss": 0.0168, "step": 16006 }, { "epoch": 4.97, "learning_rate": 3.600720144028806e-06, "loss": 0.0123, "step": 16007 }, { "epoch": 4.97, "learning_rate": 3.600320064012803e-06, "loss": 0.0141, "step": 16008 }, { "epoch": 4.97, "learning_rate": 3.5999199839967996e-06, "loss": 0.0586, "step": 16009 }, { "epoch": 4.97, "learning_rate": 3.5995199039807967e-06, "loss": 0.0065, "step": 16010 }, { "epoch": 4.97, "learning_rate": 3.5991198239647934e-06, "loss": 0.0117, "step": 16011 }, { "epoch": 4.97, "learning_rate": 3.5987197439487905e-06, "loss": 0.0169, "step": 16012 }, { "epoch": 4.97, "learning_rate": 3.5983196639327868e-06, "loss": 0.0179, "step": 16013 }, { "epoch": 4.97, "learning_rate": 3.5979195839167835e-06, "loss": 0.0255, "step": 16014 }, { "epoch": 4.97, "learning_rate": 3.5975195039007806e-06, "loss": 0.0136, "step": 16015 }, { "epoch": 4.97, "learning_rate": 3.5971194238847772e-06, "loss": 0.0328, "step": 16016 }, { "epoch": 4.97, "learning_rate": 3.5967193438687744e-06, "loss": 0.007, "step": 16017 }, { "epoch": 4.97, "learning_rate": 3.596319263852771e-06, "loss": 0.0189, "step": 16018 }, { "epoch": 4.97, "learning_rate": 3.5959191838367673e-06, "loss": 0.0145, "step": 16019 }, { "epoch": 4.97, "learning_rate": 3.5955191038207644e-06, "loss": 0.0253, "step": 16020 }, { "epoch": 4.97, "learning_rate": 3.595119023804761e-06, "loss": 0.0258, "step": 16021 }, { "epoch": 4.97, "learning_rate": 3.5947189437887582e-06, "loss": 0.0122, "step": 16022 }, { "epoch": 4.97, "learning_rate": 3.594318863772755e-06, "loss": 0.0155, "step": 16023 }, { "epoch": 4.97, "learning_rate": 3.593918783756752e-06, "loss": 0.0964, "step": 16024 }, { "epoch": 4.97, "learning_rate": 3.5935187037407483e-06, "loss": 0.0319, "step": 16025 }, { "epoch": 4.97, "learning_rate": 3.593118623724745e-06, "loss": 0.0182, "step": 16026 }, { "epoch": 4.97, "learning_rate": 3.592718543708742e-06, "loss": 0.0218, "step": 16027 }, { "epoch": 4.97, "learning_rate": 3.5923184636927388e-06, "loss": 0.0073, "step": 16028 }, { "epoch": 4.98, "learning_rate": 3.591918383676736e-06, "loss": 0.018, "step": 16029 }, { "epoch": 4.98, "learning_rate": 3.5915183036607326e-06, "loss": 0.0219, "step": 16030 }, { "epoch": 4.98, "learning_rate": 3.5911182236447297e-06, "loss": 0.023, "step": 16031 }, { "epoch": 4.98, "learning_rate": 3.590718143628726e-06, "loss": 0.0076, "step": 16032 }, { "epoch": 4.98, "learning_rate": 3.5903180636127226e-06, "loss": 0.0124, "step": 16033 }, { "epoch": 4.98, "learning_rate": 3.5899179835967197e-06, "loss": 0.0252, "step": 16034 }, { "epoch": 4.98, "learning_rate": 3.5895179035807164e-06, "loss": 0.0073, "step": 16035 }, { "epoch": 4.98, "learning_rate": 3.5891178235647135e-06, "loss": 0.0161, "step": 16036 }, { "epoch": 4.98, "learning_rate": 3.58871774354871e-06, "loss": 0.0207, "step": 16037 }, { "epoch": 4.98, "learning_rate": 3.5883176635327065e-06, "loss": 0.0095, "step": 16038 }, { "epoch": 4.98, "learning_rate": 3.5879175835167036e-06, "loss": 0.0207, "step": 16039 }, { "epoch": 4.98, "learning_rate": 3.5875175035007003e-06, "loss": 0.0661, "step": 16040 }, { "epoch": 4.98, "learning_rate": 3.5871174234846974e-06, "loss": 0.0311, "step": 16041 }, { "epoch": 4.98, "learning_rate": 3.586717343468694e-06, "loss": 0.0064, "step": 16042 }, { "epoch": 4.98, "learning_rate": 3.586317263452691e-06, "loss": 0.0147, "step": 16043 }, { "epoch": 4.98, "learning_rate": 3.5859171834366874e-06, "loss": 0.0547, "step": 16044 }, { "epoch": 4.98, "learning_rate": 3.585517103420684e-06, "loss": 0.0169, "step": 16045 }, { "epoch": 4.98, "learning_rate": 3.5851170234046812e-06, "loss": 0.03, "step": 16046 }, { "epoch": 4.98, "learning_rate": 3.584716943388678e-06, "loss": 0.0166, "step": 16047 }, { "epoch": 4.98, "learning_rate": 3.584316863372675e-06, "loss": 0.0067, "step": 16048 }, { "epoch": 4.98, "learning_rate": 3.5839167833566717e-06, "loss": 0.0086, "step": 16049 }, { "epoch": 4.98, "learning_rate": 3.583516703340669e-06, "loss": 0.005, "step": 16050 }, { "epoch": 4.98, "learning_rate": 3.583116623324665e-06, "loss": 0.023, "step": 16051 }, { "epoch": 4.98, "learning_rate": 3.5827165433086618e-06, "loss": 0.0158, "step": 16052 }, { "epoch": 4.98, "learning_rate": 3.582316463292659e-06, "loss": 0.0173, "step": 16053 }, { "epoch": 4.98, "learning_rate": 3.5819163832766556e-06, "loss": 0.0067, "step": 16054 }, { "epoch": 4.98, "learning_rate": 3.5815163032606527e-06, "loss": 0.0138, "step": 16055 }, { "epoch": 4.98, "learning_rate": 3.5811162232446494e-06, "loss": 0.0238, "step": 16056 }, { "epoch": 4.98, "learning_rate": 3.5807161432286456e-06, "loss": 0.0081, "step": 16057 }, { "epoch": 4.98, "learning_rate": 3.5803160632126427e-06, "loss": 0.0365, "step": 16058 }, { "epoch": 4.98, "learning_rate": 3.5799159831966394e-06, "loss": 0.0099, "step": 16059 }, { "epoch": 4.98, "learning_rate": 3.5795159031806365e-06, "loss": 0.0105, "step": 16060 }, { "epoch": 4.98, "learning_rate": 3.5791158231646332e-06, "loss": 0.015, "step": 16061 }, { "epoch": 4.99, "learning_rate": 3.5787157431486303e-06, "loss": 0.0095, "step": 16062 }, { "epoch": 4.99, "learning_rate": 3.5783156631326266e-06, "loss": 0.0995, "step": 16063 }, { "epoch": 4.99, "learning_rate": 3.5779155831166233e-06, "loss": 0.0073, "step": 16064 }, { "epoch": 4.99, "learning_rate": 3.5775155031006204e-06, "loss": 0.0174, "step": 16065 }, { "epoch": 4.99, "learning_rate": 3.577115423084617e-06, "loss": 0.0168, "step": 16066 }, { "epoch": 4.99, "learning_rate": 3.576715343068614e-06, "loss": 0.0562, "step": 16067 }, { "epoch": 4.99, "learning_rate": 3.576315263052611e-06, "loss": 0.0237, "step": 16068 }, { "epoch": 4.99, "learning_rate": 3.5759151830366076e-06, "loss": 0.0095, "step": 16069 }, { "epoch": 4.99, "learning_rate": 3.5755151030206042e-06, "loss": 0.0201, "step": 16070 }, { "epoch": 4.99, "learning_rate": 3.575115023004601e-06, "loss": 0.0263, "step": 16071 }, { "epoch": 4.99, "learning_rate": 3.574714942988598e-06, "loss": 0.012, "step": 16072 }, { "epoch": 4.99, "learning_rate": 3.5743148629725947e-06, "loss": 0.0326, "step": 16073 }, { "epoch": 4.99, "learning_rate": 3.573914782956592e-06, "loss": 0.0058, "step": 16074 }, { "epoch": 4.99, "learning_rate": 3.5735147029405885e-06, "loss": 0.0153, "step": 16075 }, { "epoch": 4.99, "learning_rate": 3.573114622924585e-06, "loss": 0.0216, "step": 16076 }, { "epoch": 4.99, "learning_rate": 3.572714542908582e-06, "loss": 0.008, "step": 16077 }, { "epoch": 4.99, "learning_rate": 3.5723144628925786e-06, "loss": 0.0161, "step": 16078 }, { "epoch": 4.99, "learning_rate": 3.5719143828765757e-06, "loss": 0.0174, "step": 16079 }, { "epoch": 4.99, "learning_rate": 3.5715143028605724e-06, "loss": 0.0094, "step": 16080 }, { "epoch": 4.99, "learning_rate": 3.5711142228445695e-06, "loss": 0.0172, "step": 16081 }, { "epoch": 4.99, "learning_rate": 3.5707141428285657e-06, "loss": 0.0127, "step": 16082 }, { "epoch": 4.99, "learning_rate": 3.570314062812563e-06, "loss": 0.094, "step": 16083 }, { "epoch": 4.99, "learning_rate": 3.5699139827965595e-06, "loss": 0.0318, "step": 16084 }, { "epoch": 4.99, "learning_rate": 3.5695139027805562e-06, "loss": 0.0092, "step": 16085 }, { "epoch": 4.99, "learning_rate": 3.5691138227645533e-06, "loss": 0.0062, "step": 16086 }, { "epoch": 4.99, "learning_rate": 3.56871374274855e-06, "loss": 0.0058, "step": 16087 }, { "epoch": 4.99, "learning_rate": 3.5683136627325467e-06, "loss": 0.0265, "step": 16088 }, { "epoch": 4.99, "learning_rate": 3.5679135827165434e-06, "loss": 0.0226, "step": 16089 }, { "epoch": 4.99, "learning_rate": 3.5675135027005405e-06, "loss": 0.0678, "step": 16090 }, { "epoch": 4.99, "learning_rate": 3.567113422684537e-06, "loss": 0.0092, "step": 16091 }, { "epoch": 4.99, "learning_rate": 3.566713342668534e-06, "loss": 0.0106, "step": 16092 }, { "epoch": 4.99, "learning_rate": 3.566313262652531e-06, "loss": 0.0145, "step": 16093 }, { "epoch": 5.0, "learning_rate": 3.5659131826365277e-06, "loss": 0.0713, "step": 16094 }, { "epoch": 5.0, "learning_rate": 3.5655131026205244e-06, "loss": 0.0081, "step": 16095 }, { "epoch": 5.0, "learning_rate": 3.565113022604521e-06, "loss": 0.01, "step": 16096 }, { "epoch": 5.0, "learning_rate": 3.564712942588518e-06, "loss": 0.0276, "step": 16097 }, { "epoch": 5.0, "learning_rate": 3.564312862572515e-06, "loss": 0.009, "step": 16098 }, { "epoch": 5.0, "learning_rate": 3.563912782556512e-06, "loss": 0.0144, "step": 16099 }, { "epoch": 5.0, "learning_rate": 3.5635127025405086e-06, "loss": 0.0142, "step": 16100 }, { "epoch": 5.0, "learning_rate": 3.563112622524505e-06, "loss": 0.0083, "step": 16101 }, { "epoch": 5.0, "learning_rate": 3.562712542508502e-06, "loss": 0.0339, "step": 16102 }, { "epoch": 5.0, "learning_rate": 3.5623124624924987e-06, "loss": 0.0577, "step": 16103 }, { "epoch": 5.0, "learning_rate": 3.561912382476496e-06, "loss": 0.0208, "step": 16104 }, { "epoch": 5.0, "learning_rate": 3.5615123024604925e-06, "loss": 0.0056, "step": 16105 }, { "epoch": 5.0, "learning_rate": 3.5611122224444896e-06, "loss": 0.0238, "step": 16106 }, { "epoch": 5.0, "learning_rate": 3.560712142428486e-06, "loss": 0.0167, "step": 16107 }, { "epoch": 5.0, "learning_rate": 3.5603120624124826e-06, "loss": 0.0292, "step": 16108 }, { "epoch": 5.0, "learning_rate": 3.5599119823964797e-06, "loss": 0.0174, "step": 16109 }, { "epoch": 5.0, "learning_rate": 3.5595119023804763e-06, "loss": 0.0279, "step": 16110 }, { "epoch": 5.0, "learning_rate": 3.5591118223644735e-06, "loss": 0.0051, "step": 16111 }, { "epoch": 5.0, "learning_rate": 3.55871174234847e-06, "loss": 0.0487, "step": 16112 }, { "epoch": 5.0, "learning_rate": 3.5583116623324673e-06, "loss": 0.0088, "step": 16113 }, { "epoch": 5.0, "learning_rate": 3.5579115823164635e-06, "loss": 0.0053, "step": 16114 }, { "epoch": 5.0, "learning_rate": 3.55751150230046e-06, "loss": 0.0094, "step": 16115 }, { "epoch": 5.0, "learning_rate": 3.5571114222844573e-06, "loss": 0.0037, "step": 16116 }, { "epoch": 5.0, "learning_rate": 3.556711342268454e-06, "loss": 0.0033, "step": 16117 }, { "epoch": 5.0, "learning_rate": 3.556311262252451e-06, "loss": 0.0235, "step": 16118 }, { "epoch": 5.0, "learning_rate": 3.555911182236448e-06, "loss": 0.0089, "step": 16119 }, { "epoch": 5.0, "learning_rate": 3.555511102220444e-06, "loss": 0.009, "step": 16120 }, { "epoch": 5.0, "learning_rate": 3.555111022204441e-06, "loss": 0.0096, "step": 16121 }, { "epoch": 5.0, "learning_rate": 3.554710942188438e-06, "loss": 0.0037, "step": 16122 }, { "epoch": 5.0, "learning_rate": 3.554310862172435e-06, "loss": 0.0208, "step": 16123 }, { "epoch": 5.0, "learning_rate": 3.5539107821564316e-06, "loss": 0.0021, "step": 16124 }, { "epoch": 5.0, "learning_rate": 3.5535107021404288e-06, "loss": 0.0231, "step": 16125 }, { "epoch": 5.01, "learning_rate": 3.553110622124425e-06, "loss": 0.0042, "step": 16126 }, { "epoch": 5.01, "learning_rate": 3.5527105421084217e-06, "loss": 0.0028, "step": 16127 }, { "epoch": 5.01, "learning_rate": 3.552310462092419e-06, "loss": 0.0068, "step": 16128 }, { "epoch": 5.01, "learning_rate": 3.5519103820764155e-06, "loss": 0.0391, "step": 16129 }, { "epoch": 5.01, "learning_rate": 3.5515103020604126e-06, "loss": 0.0216, "step": 16130 }, { "epoch": 5.01, "learning_rate": 3.5511102220444093e-06, "loss": 0.0039, "step": 16131 }, { "epoch": 5.01, "learning_rate": 3.5507101420284064e-06, "loss": 0.0133, "step": 16132 }, { "epoch": 5.01, "learning_rate": 3.5503100620124027e-06, "loss": 0.0166, "step": 16133 }, { "epoch": 5.01, "learning_rate": 3.5499099819963994e-06, "loss": 0.0059, "step": 16134 }, { "epoch": 5.01, "learning_rate": 3.5495099019803965e-06, "loss": 0.0095, "step": 16135 }, { "epoch": 5.01, "learning_rate": 3.549109821964393e-06, "loss": 0.0737, "step": 16136 }, { "epoch": 5.01, "learning_rate": 3.5487097419483903e-06, "loss": 0.0096, "step": 16137 }, { "epoch": 5.01, "learning_rate": 3.548309661932387e-06, "loss": 0.0029, "step": 16138 }, { "epoch": 5.01, "learning_rate": 3.5479095819163832e-06, "loss": 0.0079, "step": 16139 }, { "epoch": 5.01, "learning_rate": 3.5475095019003803e-06, "loss": 0.0086, "step": 16140 }, { "epoch": 5.01, "learning_rate": 3.547109421884377e-06, "loss": 0.007, "step": 16141 }, { "epoch": 5.01, "learning_rate": 3.546709341868374e-06, "loss": 0.0037, "step": 16142 }, { "epoch": 5.01, "learning_rate": 3.546309261852371e-06, "loss": 0.0045, "step": 16143 }, { "epoch": 5.01, "learning_rate": 3.545909181836368e-06, "loss": 0.0038, "step": 16144 }, { "epoch": 5.01, "learning_rate": 3.545509101820364e-06, "loss": 0.0026, "step": 16145 }, { "epoch": 5.01, "learning_rate": 3.545109021804361e-06, "loss": 0.0042, "step": 16146 }, { "epoch": 5.01, "learning_rate": 3.544708941788358e-06, "loss": 0.0026, "step": 16147 }, { "epoch": 5.01, "learning_rate": 3.5443088617723547e-06, "loss": 0.0068, "step": 16148 }, { "epoch": 5.01, "learning_rate": 3.5439087817563518e-06, "loss": 0.0049, "step": 16149 }, { "epoch": 5.01, "learning_rate": 3.5435087017403485e-06, "loss": 0.0049, "step": 16150 }, { "epoch": 5.01, "learning_rate": 3.5431086217243447e-06, "loss": 0.0038, "step": 16151 }, { "epoch": 5.01, "learning_rate": 3.542708541708342e-06, "loss": 0.0074, "step": 16152 }, { "epoch": 5.01, "learning_rate": 3.5423084616923385e-06, "loss": 0.0029, "step": 16153 }, { "epoch": 5.01, "learning_rate": 3.5419083816763356e-06, "loss": 0.0133, "step": 16154 }, { "epoch": 5.01, "learning_rate": 3.5415083016603323e-06, "loss": 0.0125, "step": 16155 }, { "epoch": 5.01, "learning_rate": 3.5411082216443294e-06, "loss": 0.0372, "step": 16156 }, { "epoch": 5.01, "learning_rate": 3.540708141628326e-06, "loss": 0.0019, "step": 16157 }, { "epoch": 5.02, "learning_rate": 3.5403080616123224e-06, "loss": 0.0117, "step": 16158 }, { "epoch": 5.02, "learning_rate": 3.5399079815963195e-06, "loss": 0.005, "step": 16159 }, { "epoch": 5.02, "learning_rate": 3.539507901580316e-06, "loss": 0.0102, "step": 16160 }, { "epoch": 5.02, "learning_rate": 3.5391078215643133e-06, "loss": 0.0032, "step": 16161 }, { "epoch": 5.02, "learning_rate": 3.53870774154831e-06, "loss": 0.0029, "step": 16162 }, { "epoch": 5.02, "learning_rate": 3.538307661532307e-06, "loss": 0.0079, "step": 16163 }, { "epoch": 5.02, "learning_rate": 3.5379075815163033e-06, "loss": 0.005, "step": 16164 }, { "epoch": 5.02, "learning_rate": 3.5375075015003e-06, "loss": 0.0065, "step": 16165 }, { "epoch": 5.02, "learning_rate": 3.537107421484297e-06, "loss": 0.0075, "step": 16166 }, { "epoch": 5.02, "learning_rate": 3.536707341468294e-06, "loss": 0.0028, "step": 16167 }, { "epoch": 5.02, "learning_rate": 3.536307261452291e-06, "loss": 0.0056, "step": 16168 }, { "epoch": 5.02, "learning_rate": 3.5359071814362876e-06, "loss": 0.0027, "step": 16169 }, { "epoch": 5.02, "learning_rate": 3.535507101420284e-06, "loss": 0.0056, "step": 16170 }, { "epoch": 5.02, "learning_rate": 3.535107021404281e-06, "loss": 0.0038, "step": 16171 }, { "epoch": 5.02, "learning_rate": 3.5347069413882777e-06, "loss": 0.0075, "step": 16172 }, { "epoch": 5.02, "learning_rate": 3.5343068613722748e-06, "loss": 0.0034, "step": 16173 }, { "epoch": 5.02, "learning_rate": 3.5339067813562715e-06, "loss": 0.0052, "step": 16174 }, { "epoch": 5.02, "learning_rate": 3.5335067013402686e-06, "loss": 0.0052, "step": 16175 }, { "epoch": 5.02, "learning_rate": 3.5331066213242653e-06, "loss": 0.0163, "step": 16176 }, { "epoch": 5.02, "learning_rate": 3.5327065413082615e-06, "loss": 0.003, "step": 16177 }, { "epoch": 5.02, "learning_rate": 3.5323064612922586e-06, "loss": 0.0027, "step": 16178 }, { "epoch": 5.02, "learning_rate": 3.5319063812762553e-06, "loss": 0.0033, "step": 16179 }, { "epoch": 5.02, "learning_rate": 3.5315063012602524e-06, "loss": 0.0056, "step": 16180 }, { "epoch": 5.02, "learning_rate": 3.531106221244249e-06, "loss": 0.0058, "step": 16181 }, { "epoch": 5.02, "learning_rate": 3.5307061412282462e-06, "loss": 0.0044, "step": 16182 }, { "epoch": 5.02, "learning_rate": 3.5303060612122425e-06, "loss": 0.0041, "step": 16183 }, { "epoch": 5.02, "learning_rate": 3.5299059811962396e-06, "loss": 0.0037, "step": 16184 }, { "epoch": 5.02, "learning_rate": 3.5295059011802363e-06, "loss": 0.0122, "step": 16185 }, { "epoch": 5.02, "learning_rate": 3.529105821164233e-06, "loss": 0.0153, "step": 16186 }, { "epoch": 5.02, "learning_rate": 3.52870574114823e-06, "loss": 0.0031, "step": 16187 }, { "epoch": 5.02, "learning_rate": 3.5283056611322268e-06, "loss": 0.0052, "step": 16188 }, { "epoch": 5.02, "learning_rate": 3.5279055811162235e-06, "loss": 0.0028, "step": 16189 }, { "epoch": 5.03, "learning_rate": 3.52750550110022e-06, "loss": 0.0025, "step": 16190 }, { "epoch": 5.03, "learning_rate": 3.5271054210842173e-06, "loss": 0.004, "step": 16191 }, { "epoch": 5.03, "learning_rate": 3.526705341068214e-06, "loss": 0.0189, "step": 16192 }, { "epoch": 5.03, "learning_rate": 3.5263052610522106e-06, "loss": 0.0056, "step": 16193 }, { "epoch": 5.03, "learning_rate": 3.5259051810362077e-06, "loss": 0.0031, "step": 16194 }, { "epoch": 5.03, "learning_rate": 3.5255051010202044e-06, "loss": 0.0153, "step": 16195 }, { "epoch": 5.03, "learning_rate": 3.525105021004201e-06, "loss": 0.0059, "step": 16196 }, { "epoch": 5.03, "learning_rate": 3.524704940988198e-06, "loss": 0.0029, "step": 16197 }, { "epoch": 5.03, "learning_rate": 3.524304860972195e-06, "loss": 0.0744, "step": 16198 }, { "epoch": 5.03, "learning_rate": 3.5239047809561916e-06, "loss": 0.0345, "step": 16199 }, { "epoch": 5.03, "learning_rate": 3.5235047009401883e-06, "loss": 0.0458, "step": 16200 }, { "epoch": 5.03, "learning_rate": 3.5231046209241854e-06, "loss": 0.0025, "step": 16201 }, { "epoch": 5.03, "learning_rate": 3.5227045409081816e-06, "loss": 0.0073, "step": 16202 }, { "epoch": 5.03, "learning_rate": 3.5223044608921788e-06, "loss": 0.0035, "step": 16203 }, { "epoch": 5.03, "learning_rate": 3.5219043808761754e-06, "loss": 0.0015, "step": 16204 }, { "epoch": 5.03, "learning_rate": 3.5215043008601726e-06, "loss": 0.0068, "step": 16205 }, { "epoch": 5.03, "learning_rate": 3.5211042208441692e-06, "loss": 0.0026, "step": 16206 }, { "epoch": 5.03, "learning_rate": 3.5207041408281664e-06, "loss": 0.0072, "step": 16207 }, { "epoch": 5.03, "learning_rate": 3.5203040608121626e-06, "loss": 0.0031, "step": 16208 }, { "epoch": 5.03, "learning_rate": 3.5199039807961593e-06, "loss": 0.0149, "step": 16209 }, { "epoch": 5.03, "learning_rate": 3.5195039007801564e-06, "loss": 0.0022, "step": 16210 }, { "epoch": 5.03, "learning_rate": 3.519103820764153e-06, "loss": 0.0024, "step": 16211 }, { "epoch": 5.03, "learning_rate": 3.51870374074815e-06, "loss": 0.0038, "step": 16212 }, { "epoch": 5.03, "learning_rate": 3.518303660732147e-06, "loss": 0.0037, "step": 16213 }, { "epoch": 5.03, "learning_rate": 3.517903580716143e-06, "loss": 0.0132, "step": 16214 }, { "epoch": 5.03, "learning_rate": 3.5175035007001403e-06, "loss": 0.0175, "step": 16215 }, { "epoch": 5.03, "learning_rate": 3.517103420684137e-06, "loss": 0.0297, "step": 16216 }, { "epoch": 5.03, "learning_rate": 3.516703340668134e-06, "loss": 0.0109, "step": 16217 }, { "epoch": 5.03, "learning_rate": 3.5163032606521307e-06, "loss": 0.0096, "step": 16218 }, { "epoch": 5.03, "learning_rate": 3.515903180636128e-06, "loss": 0.0075, "step": 16219 }, { "epoch": 5.03, "learning_rate": 3.5155031006201245e-06, "loss": 0.0032, "step": 16220 }, { "epoch": 5.03, "learning_rate": 3.515103020604121e-06, "loss": 0.0022, "step": 16221 }, { "epoch": 5.03, "learning_rate": 3.514702940588118e-06, "loss": 0.0046, "step": 16222 }, { "epoch": 5.04, "learning_rate": 3.5143028605721146e-06, "loss": 0.0044, "step": 16223 }, { "epoch": 5.04, "learning_rate": 3.5139027805561117e-06, "loss": 0.0175, "step": 16224 }, { "epoch": 5.04, "learning_rate": 3.5135027005401084e-06, "loss": 0.003, "step": 16225 }, { "epoch": 5.04, "learning_rate": 3.5131026205241055e-06, "loss": 0.0568, "step": 16226 }, { "epoch": 5.04, "learning_rate": 3.5127025405081018e-06, "loss": 0.0076, "step": 16227 }, { "epoch": 5.04, "learning_rate": 3.5123024604920985e-06, "loss": 0.0045, "step": 16228 }, { "epoch": 5.04, "learning_rate": 3.5119023804760956e-06, "loss": 0.0047, "step": 16229 }, { "epoch": 5.04, "learning_rate": 3.5115023004600923e-06, "loss": 0.0036, "step": 16230 }, { "epoch": 5.04, "learning_rate": 3.5111022204440894e-06, "loss": 0.0054, "step": 16231 }, { "epoch": 5.04, "learning_rate": 3.510702140428086e-06, "loss": 0.0113, "step": 16232 }, { "epoch": 5.04, "learning_rate": 3.5103020604120823e-06, "loss": 0.0066, "step": 16233 }, { "epoch": 5.04, "learning_rate": 3.5099019803960794e-06, "loss": 0.0126, "step": 16234 }, { "epoch": 5.04, "learning_rate": 3.509501900380076e-06, "loss": 0.0082, "step": 16235 }, { "epoch": 5.04, "learning_rate": 3.5091018203640732e-06, "loss": 0.0076, "step": 16236 }, { "epoch": 5.04, "learning_rate": 3.50870174034807e-06, "loss": 0.008, "step": 16237 }, { "epoch": 5.04, "learning_rate": 3.508301660332067e-06, "loss": 0.0083, "step": 16238 }, { "epoch": 5.04, "learning_rate": 3.5079015803160637e-06, "loss": 0.0115, "step": 16239 }, { "epoch": 5.04, "learning_rate": 3.50750150030006e-06, "loss": 0.0193, "step": 16240 }, { "epoch": 5.04, "learning_rate": 3.507101420284057e-06, "loss": 0.0044, "step": 16241 }, { "epoch": 5.04, "learning_rate": 3.5067013402680538e-06, "loss": 0.0026, "step": 16242 }, { "epoch": 5.04, "learning_rate": 3.506301260252051e-06, "loss": 0.0032, "step": 16243 }, { "epoch": 5.04, "learning_rate": 3.5059011802360476e-06, "loss": 0.0481, "step": 16244 }, { "epoch": 5.04, "learning_rate": 3.5055011002200447e-06, "loss": 0.007, "step": 16245 }, { "epoch": 5.04, "learning_rate": 3.505101020204041e-06, "loss": 0.0818, "step": 16246 }, { "epoch": 5.04, "learning_rate": 3.5047009401880376e-06, "loss": 0.01, "step": 16247 }, { "epoch": 5.04, "learning_rate": 3.5043008601720347e-06, "loss": 0.0021, "step": 16248 }, { "epoch": 5.04, "learning_rate": 3.5039007801560314e-06, "loss": 0.0099, "step": 16249 }, { "epoch": 5.04, "learning_rate": 3.5035007001400285e-06, "loss": 0.0041, "step": 16250 }, { "epoch": 5.04, "learning_rate": 3.503100620124025e-06, "loss": 0.0147, "step": 16251 }, { "epoch": 5.04, "learning_rate": 3.5027005401080215e-06, "loss": 0.0477, "step": 16252 }, { "epoch": 5.04, "learning_rate": 3.5023004600920186e-06, "loss": 0.0228, "step": 16253 }, { "epoch": 5.04, "learning_rate": 3.5019003800760153e-06, "loss": 0.0035, "step": 16254 }, { "epoch": 5.05, "learning_rate": 3.5015003000600124e-06, "loss": 0.0026, "step": 16255 }, { "epoch": 5.05, "learning_rate": 3.501100220044009e-06, "loss": 0.0044, "step": 16256 }, { "epoch": 5.05, "learning_rate": 3.500700140028006e-06, "loss": 0.009, "step": 16257 }, { "epoch": 5.05, "learning_rate": 3.500300060012003e-06, "loss": 0.0167, "step": 16258 }, { "epoch": 5.05, "learning_rate": 3.499899979995999e-06, "loss": 0.0095, "step": 16259 }, { "epoch": 5.05, "learning_rate": 3.4994998999799962e-06, "loss": 0.0334, "step": 16260 }, { "epoch": 5.05, "learning_rate": 3.499099819963993e-06, "loss": 0.0072, "step": 16261 }, { "epoch": 5.05, "learning_rate": 3.49869973994799e-06, "loss": 0.0088, "step": 16262 }, { "epoch": 5.05, "learning_rate": 3.4982996599319867e-06, "loss": 0.0043, "step": 16263 }, { "epoch": 5.05, "learning_rate": 3.497899579915984e-06, "loss": 0.0056, "step": 16264 }, { "epoch": 5.05, "learning_rate": 3.49749949989998e-06, "loss": 0.0064, "step": 16265 }, { "epoch": 5.05, "learning_rate": 3.4970994198839768e-06, "loss": 0.0138, "step": 16266 }, { "epoch": 5.05, "learning_rate": 3.496699339867974e-06, "loss": 0.0072, "step": 16267 }, { "epoch": 5.05, "learning_rate": 3.4962992598519706e-06, "loss": 0.0078, "step": 16268 }, { "epoch": 5.05, "learning_rate": 3.4958991798359677e-06, "loss": 0.0073, "step": 16269 }, { "epoch": 5.05, "learning_rate": 3.4954990998199644e-06, "loss": 0.0123, "step": 16270 }, { "epoch": 5.05, "learning_rate": 3.4950990198039606e-06, "loss": 0.0037, "step": 16271 }, { "epoch": 5.05, "learning_rate": 3.4946989397879577e-06, "loss": 0.0515, "step": 16272 }, { "epoch": 5.05, "learning_rate": 3.4942988597719544e-06, "loss": 0.0046, "step": 16273 }, { "epoch": 5.05, "learning_rate": 3.4938987797559515e-06, "loss": 0.0023, "step": 16274 }, { "epoch": 5.05, "learning_rate": 3.4934986997399482e-06, "loss": 0.0089, "step": 16275 }, { "epoch": 5.05, "learning_rate": 3.4930986197239453e-06, "loss": 0.0042, "step": 16276 }, { "epoch": 5.05, "learning_rate": 3.492698539707942e-06, "loss": 0.0125, "step": 16277 }, { "epoch": 5.05, "learning_rate": 3.4922984596919383e-06, "loss": 0.006, "step": 16278 }, { "epoch": 5.05, "learning_rate": 3.4918983796759354e-06, "loss": 0.0038, "step": 16279 }, { "epoch": 5.05, "learning_rate": 3.491498299659932e-06, "loss": 0.0128, "step": 16280 }, { "epoch": 5.05, "learning_rate": 3.491098219643929e-06, "loss": 0.0114, "step": 16281 }, { "epoch": 5.05, "learning_rate": 3.490698139627926e-06, "loss": 0.0032, "step": 16282 }, { "epoch": 5.05, "learning_rate": 3.490298059611923e-06, "loss": 0.0039, "step": 16283 }, { "epoch": 5.05, "learning_rate": 3.4898979795959192e-06, "loss": 0.0053, "step": 16284 }, { "epoch": 5.05, "learning_rate": 3.489497899579916e-06, "loss": 0.0041, "step": 16285 }, { "epoch": 5.05, "learning_rate": 3.489097819563913e-06, "loss": 0.0133, "step": 16286 }, { "epoch": 5.06, "learning_rate": 3.4886977395479097e-06, "loss": 0.0014, "step": 16287 }, { "epoch": 5.06, "learning_rate": 3.488297659531907e-06, "loss": 0.0525, "step": 16288 }, { "epoch": 5.06, "learning_rate": 3.4878975795159035e-06, "loss": 0.0042, "step": 16289 }, { "epoch": 5.06, "learning_rate": 3.4874974994999e-06, "loss": 0.0287, "step": 16290 }, { "epoch": 5.06, "learning_rate": 3.487097419483897e-06, "loss": 0.0044, "step": 16291 }, { "epoch": 5.06, "learning_rate": 3.486697339467894e-06, "loss": 0.0056, "step": 16292 }, { "epoch": 5.06, "learning_rate": 3.4862972594518907e-06, "loss": 0.0035, "step": 16293 }, { "epoch": 5.06, "learning_rate": 3.4858971794358874e-06, "loss": 0.0166, "step": 16294 }, { "epoch": 5.06, "learning_rate": 3.4854970994198845e-06, "loss": 0.0051, "step": 16295 }, { "epoch": 5.06, "learning_rate": 3.4850970194038807e-06, "loss": 0.0186, "step": 16296 }, { "epoch": 5.06, "learning_rate": 3.484696939387878e-06, "loss": 0.0119, "step": 16297 }, { "epoch": 5.06, "learning_rate": 3.4842968593718745e-06, "loss": 0.0104, "step": 16298 }, { "epoch": 5.06, "learning_rate": 3.4838967793558717e-06, "loss": 0.0054, "step": 16299 }, { "epoch": 5.06, "learning_rate": 3.4834966993398683e-06, "loss": 0.0026, "step": 16300 }, { "epoch": 5.06, "learning_rate": 3.483096619323865e-06, "loss": 0.0026, "step": 16301 }, { "epoch": 5.06, "learning_rate": 3.482696539307862e-06, "loss": 0.0087, "step": 16302 }, { "epoch": 5.06, "learning_rate": 3.4822964592918584e-06, "loss": 0.0027, "step": 16303 }, { "epoch": 5.06, "learning_rate": 3.4818963792758555e-06, "loss": 0.0048, "step": 16304 }, { "epoch": 5.06, "learning_rate": 3.481496299259852e-06, "loss": 0.0283, "step": 16305 }, { "epoch": 5.06, "learning_rate": 3.4810962192438493e-06, "loss": 0.0052, "step": 16306 }, { "epoch": 5.06, "learning_rate": 3.480696139227846e-06, "loss": 0.0159, "step": 16307 }, { "epoch": 5.06, "learning_rate": 3.4802960592118427e-06, "loss": 0.0019, "step": 16308 }, { "epoch": 5.06, "learning_rate": 3.4798959791958394e-06, "loss": 0.0134, "step": 16309 }, { "epoch": 5.06, "learning_rate": 3.479495899179836e-06, "loss": 0.0748, "step": 16310 }, { "epoch": 5.06, "learning_rate": 3.479095819163833e-06, "loss": 0.0037, "step": 16311 }, { "epoch": 5.06, "learning_rate": 3.47869573914783e-06, "loss": 0.0049, "step": 16312 }, { "epoch": 5.06, "learning_rate": 3.478295659131827e-06, "loss": 0.0017, "step": 16313 }, { "epoch": 5.06, "learning_rate": 3.4778955791158236e-06, "loss": 0.0029, "step": 16314 }, { "epoch": 5.06, "learning_rate": 3.47749549909982e-06, "loss": 0.0071, "step": 16315 }, { "epoch": 5.06, "learning_rate": 3.477095419083817e-06, "loss": 0.0055, "step": 16316 }, { "epoch": 5.06, "learning_rate": 3.4766953390678137e-06, "loss": 0.0103, "step": 16317 }, { "epoch": 5.06, "learning_rate": 3.476295259051811e-06, "loss": 0.0055, "step": 16318 }, { "epoch": 5.07, "learning_rate": 3.4758951790358075e-06, "loss": 0.0044, "step": 16319 }, { "epoch": 5.07, "learning_rate": 3.4754950990198046e-06, "loss": 0.0105, "step": 16320 }, { "epoch": 5.07, "learning_rate": 3.4750950190038013e-06, "loss": 0.0073, "step": 16321 }, { "epoch": 5.07, "learning_rate": 3.4746949389877976e-06, "loss": 0.0077, "step": 16322 }, { "epoch": 5.07, "learning_rate": 3.4742948589717947e-06, "loss": 0.012, "step": 16323 }, { "epoch": 5.07, "learning_rate": 3.4738947789557914e-06, "loss": 0.0056, "step": 16324 }, { "epoch": 5.07, "learning_rate": 3.4734946989397885e-06, "loss": 0.0465, "step": 16325 }, { "epoch": 5.07, "learning_rate": 3.473094618923785e-06, "loss": 0.0144, "step": 16326 }, { "epoch": 5.07, "learning_rate": 3.4726945389077823e-06, "loss": 0.0065, "step": 16327 }, { "epoch": 5.07, "learning_rate": 3.4722944588917785e-06, "loss": 0.0084, "step": 16328 }, { "epoch": 5.07, "learning_rate": 3.471894378875775e-06, "loss": 0.0156, "step": 16329 }, { "epoch": 5.07, "learning_rate": 3.4714942988597723e-06, "loss": 0.0556, "step": 16330 }, { "epoch": 5.07, "learning_rate": 3.471094218843769e-06, "loss": 0.0052, "step": 16331 }, { "epoch": 5.07, "learning_rate": 3.470694138827766e-06, "loss": 0.0071, "step": 16332 }, { "epoch": 5.07, "learning_rate": 3.470294058811763e-06, "loss": 0.0142, "step": 16333 }, { "epoch": 5.07, "learning_rate": 3.469893978795759e-06, "loss": 0.0061, "step": 16334 }, { "epoch": 5.07, "learning_rate": 3.469493898779756e-06, "loss": 0.0118, "step": 16335 }, { "epoch": 5.07, "learning_rate": 3.469093818763753e-06, "loss": 0.0103, "step": 16336 }, { "epoch": 5.07, "learning_rate": 3.46869373874775e-06, "loss": 0.0065, "step": 16337 }, { "epoch": 5.07, "learning_rate": 3.4682936587317467e-06, "loss": 0.0052, "step": 16338 }, { "epoch": 5.07, "learning_rate": 3.4678935787157438e-06, "loss": 0.0033, "step": 16339 }, { "epoch": 5.07, "learning_rate": 3.4674934986997404e-06, "loss": 0.0056, "step": 16340 }, { "epoch": 5.07, "learning_rate": 3.4670934186837367e-06, "loss": 0.0139, "step": 16341 }, { "epoch": 5.07, "learning_rate": 3.466693338667734e-06, "loss": 0.0178, "step": 16342 }, { "epoch": 5.07, "learning_rate": 3.4662932586517305e-06, "loss": 0.0092, "step": 16343 }, { "epoch": 5.07, "learning_rate": 3.4658931786357276e-06, "loss": 0.0567, "step": 16344 }, { "epoch": 5.07, "learning_rate": 3.4654930986197243e-06, "loss": 0.0166, "step": 16345 }, { "epoch": 5.07, "learning_rate": 3.4650930186037214e-06, "loss": 0.0058, "step": 16346 }, { "epoch": 5.07, "learning_rate": 3.4646929385877177e-06, "loss": 0.0017, "step": 16347 }, { "epoch": 5.07, "learning_rate": 3.4642928585717144e-06, "loss": 0.0043, "step": 16348 }, { "epoch": 5.07, "learning_rate": 3.4638927785557115e-06, "loss": 0.0058, "step": 16349 }, { "epoch": 5.07, "learning_rate": 3.463492698539708e-06, "loss": 0.0043, "step": 16350 }, { "epoch": 5.07, "learning_rate": 3.4630926185237053e-06, "loss": 0.0044, "step": 16351 }, { "epoch": 5.08, "learning_rate": 3.462692538507702e-06, "loss": 0.0048, "step": 16352 }, { "epoch": 5.08, "learning_rate": 3.4622924584916982e-06, "loss": 0.0081, "step": 16353 }, { "epoch": 5.08, "learning_rate": 3.4618923784756953e-06, "loss": 0.0747, "step": 16354 }, { "epoch": 5.08, "learning_rate": 3.461492298459692e-06, "loss": 0.0159, "step": 16355 }, { "epoch": 5.08, "learning_rate": 3.461092218443689e-06, "loss": 0.0023, "step": 16356 }, { "epoch": 5.08, "learning_rate": 3.460692138427686e-06, "loss": 0.0055, "step": 16357 }, { "epoch": 5.08, "learning_rate": 3.460292058411683e-06, "loss": 0.0111, "step": 16358 }, { "epoch": 5.08, "learning_rate": 3.4598919783956796e-06, "loss": 0.0194, "step": 16359 }, { "epoch": 5.08, "learning_rate": 3.459491898379676e-06, "loss": 0.003, "step": 16360 }, { "epoch": 5.08, "learning_rate": 3.459091818363673e-06, "loss": 0.0113, "step": 16361 }, { "epoch": 5.08, "learning_rate": 3.4586917383476697e-06, "loss": 0.0212, "step": 16362 }, { "epoch": 5.08, "learning_rate": 3.4582916583316668e-06, "loss": 0.008, "step": 16363 }, { "epoch": 5.08, "learning_rate": 3.4578915783156635e-06, "loss": 0.0986, "step": 16364 }, { "epoch": 5.08, "learning_rate": 3.4574914982996606e-06, "loss": 0.0066, "step": 16365 }, { "epoch": 5.08, "learning_rate": 3.457091418283657e-06, "loss": 0.0043, "step": 16366 }, { "epoch": 5.08, "learning_rate": 3.4566913382676535e-06, "loss": 0.0056, "step": 16367 }, { "epoch": 5.08, "learning_rate": 3.4562912582516506e-06, "loss": 0.002, "step": 16368 }, { "epoch": 5.08, "learning_rate": 3.4558911782356473e-06, "loss": 0.0094, "step": 16369 }, { "epoch": 5.08, "learning_rate": 3.4554910982196444e-06, "loss": 0.0047, "step": 16370 }, { "epoch": 5.08, "learning_rate": 3.455091018203641e-06, "loss": 0.0098, "step": 16371 }, { "epoch": 5.08, "learning_rate": 3.4546909381876374e-06, "loss": 0.0143, "step": 16372 }, { "epoch": 5.08, "learning_rate": 3.4542908581716345e-06, "loss": 0.007, "step": 16373 }, { "epoch": 5.08, "learning_rate": 3.453890778155631e-06, "loss": 0.008, "step": 16374 }, { "epoch": 5.08, "learning_rate": 3.4534906981396283e-06, "loss": 0.0732, "step": 16375 }, { "epoch": 5.08, "learning_rate": 3.453090618123625e-06, "loss": 0.019, "step": 16376 }, { "epoch": 5.08, "learning_rate": 3.452690538107622e-06, "loss": 0.0094, "step": 16377 }, { "epoch": 5.08, "learning_rate": 3.4522904580916183e-06, "loss": 0.0061, "step": 16378 }, { "epoch": 5.08, "learning_rate": 3.451890378075615e-06, "loss": 0.0227, "step": 16379 }, { "epoch": 5.08, "learning_rate": 3.451490298059612e-06, "loss": 0.0089, "step": 16380 }, { "epoch": 5.08, "learning_rate": 3.451090218043609e-06, "loss": 0.015, "step": 16381 }, { "epoch": 5.08, "learning_rate": 3.450690138027606e-06, "loss": 0.0053, "step": 16382 }, { "epoch": 5.08, "learning_rate": 3.4502900580116026e-06, "loss": 0.0164, "step": 16383 }, { "epoch": 5.09, "learning_rate": 3.4498899779955997e-06, "loss": 0.0047, "step": 16384 }, { "epoch": 5.09, "learning_rate": 3.449489897979596e-06, "loss": 0.0089, "step": 16385 }, { "epoch": 5.09, "learning_rate": 3.4490898179635927e-06, "loss": 0.0084, "step": 16386 }, { "epoch": 5.09, "learning_rate": 3.4486897379475898e-06, "loss": 0.0068, "step": 16387 }, { "epoch": 5.09, "learning_rate": 3.4482896579315865e-06, "loss": 0.0132, "step": 16388 }, { "epoch": 5.09, "learning_rate": 3.4478895779155836e-06, "loss": 0.0201, "step": 16389 }, { "epoch": 5.09, "learning_rate": 3.4474894978995803e-06, "loss": 0.0053, "step": 16390 }, { "epoch": 5.09, "learning_rate": 3.447089417883577e-06, "loss": 0.0161, "step": 16391 }, { "epoch": 5.09, "learning_rate": 3.4466893378675736e-06, "loss": 0.0032, "step": 16392 }, { "epoch": 5.09, "learning_rate": 3.4462892578515703e-06, "loss": 0.0064, "step": 16393 }, { "epoch": 5.09, "learning_rate": 3.4458891778355674e-06, "loss": 0.0043, "step": 16394 }, { "epoch": 5.09, "learning_rate": 3.445489097819564e-06, "loss": 0.0027, "step": 16395 }, { "epoch": 5.09, "learning_rate": 3.4450890178035612e-06, "loss": 0.0065, "step": 16396 }, { "epoch": 5.09, "learning_rate": 3.4446889377875575e-06, "loss": 0.0026, "step": 16397 }, { "epoch": 5.09, "learning_rate": 3.4442888577715546e-06, "loss": 0.0045, "step": 16398 }, { "epoch": 5.09, "learning_rate": 3.4438887777555513e-06, "loss": 0.0033, "step": 16399 }, { "epoch": 5.09, "learning_rate": 3.4434886977395484e-06, "loss": 0.0034, "step": 16400 }, { "epoch": 5.09, "learning_rate": 3.443088617723545e-06, "loss": 0.0026, "step": 16401 }, { "epoch": 5.09, "learning_rate": 3.4426885377075418e-06, "loss": 0.0114, "step": 16402 }, { "epoch": 5.09, "learning_rate": 3.442288457691539e-06, "loss": 0.0013, "step": 16403 }, { "epoch": 5.09, "learning_rate": 3.441888377675535e-06, "loss": 0.0047, "step": 16404 }, { "epoch": 5.09, "learning_rate": 3.4414882976595323e-06, "loss": 0.0156, "step": 16405 }, { "epoch": 5.09, "learning_rate": 3.441088217643529e-06, "loss": 0.0026, "step": 16406 }, { "epoch": 5.09, "learning_rate": 3.440688137627526e-06, "loss": 0.0018, "step": 16407 }, { "epoch": 5.09, "learning_rate": 3.4402880576115227e-06, "loss": 0.0055, "step": 16408 }, { "epoch": 5.09, "learning_rate": 3.4398879775955194e-06, "loss": 0.0154, "step": 16409 }, { "epoch": 5.09, "learning_rate": 3.439487897579516e-06, "loss": 0.0063, "step": 16410 }, { "epoch": 5.09, "learning_rate": 3.439087817563513e-06, "loss": 0.0061, "step": 16411 }, { "epoch": 5.09, "learning_rate": 3.43868773754751e-06, "loss": 0.006, "step": 16412 }, { "epoch": 5.09, "learning_rate": 3.4382876575315066e-06, "loss": 0.0416, "step": 16413 }, { "epoch": 5.09, "learning_rate": 3.4378875775155037e-06, "loss": 0.0118, "step": 16414 }, { "epoch": 5.09, "learning_rate": 3.4374874974995004e-06, "loss": 0.0596, "step": 16415 }, { "epoch": 5.1, "learning_rate": 3.4370874174834967e-06, "loss": 0.005, "step": 16416 }, { "epoch": 5.1, "learning_rate": 3.4366873374674938e-06, "loss": 0.003, "step": 16417 }, { "epoch": 5.1, "learning_rate": 3.4362872574514904e-06, "loss": 0.0087, "step": 16418 }, { "epoch": 5.1, "learning_rate": 3.4358871774354876e-06, "loss": 0.0153, "step": 16419 }, { "epoch": 5.1, "learning_rate": 3.4354870974194842e-06, "loss": 0.015, "step": 16420 }, { "epoch": 5.1, "learning_rate": 3.4350870174034814e-06, "loss": 0.005, "step": 16421 }, { "epoch": 5.1, "learning_rate": 3.434686937387478e-06, "loss": 0.0148, "step": 16422 }, { "epoch": 5.1, "learning_rate": 3.4342868573714743e-06, "loss": 0.0037, "step": 16423 }, { "epoch": 5.1, "learning_rate": 3.4338867773554714e-06, "loss": 0.0016, "step": 16424 }, { "epoch": 5.1, "learning_rate": 3.433486697339468e-06, "loss": 0.0369, "step": 16425 }, { "epoch": 5.1, "learning_rate": 3.433086617323465e-06, "loss": 0.0347, "step": 16426 }, { "epoch": 5.1, "learning_rate": 3.432686537307462e-06, "loss": 0.0516, "step": 16427 }, { "epoch": 5.1, "learning_rate": 3.432286457291459e-06, "loss": 0.0022, "step": 16428 }, { "epoch": 5.1, "learning_rate": 3.4318863772754553e-06, "loss": 0.0052, "step": 16429 }, { "epoch": 5.1, "learning_rate": 3.431486297259452e-06, "loss": 0.0786, "step": 16430 }, { "epoch": 5.1, "learning_rate": 3.431086217243449e-06, "loss": 0.0049, "step": 16431 }, { "epoch": 5.1, "learning_rate": 3.4306861372274458e-06, "loss": 0.0025, "step": 16432 }, { "epoch": 5.1, "learning_rate": 3.430286057211443e-06, "loss": 0.0187, "step": 16433 }, { "epoch": 5.1, "learning_rate": 3.4298859771954395e-06, "loss": 0.0094, "step": 16434 }, { "epoch": 5.1, "learning_rate": 3.429485897179436e-06, "loss": 0.0178, "step": 16435 }, { "epoch": 5.1, "learning_rate": 3.429085817163433e-06, "loss": 0.0036, "step": 16436 }, { "epoch": 5.1, "learning_rate": 3.4286857371474296e-06, "loss": 0.0071, "step": 16437 }, { "epoch": 5.1, "learning_rate": 3.4282856571314267e-06, "loss": 0.0081, "step": 16438 }, { "epoch": 5.1, "learning_rate": 3.4278855771154234e-06, "loss": 0.0087, "step": 16439 }, { "epoch": 5.1, "learning_rate": 3.4274854970994205e-06, "loss": 0.0028, "step": 16440 }, { "epoch": 5.1, "learning_rate": 3.4270854170834168e-06, "loss": 0.002, "step": 16441 }, { "epoch": 5.1, "learning_rate": 3.4266853370674135e-06, "loss": 0.0168, "step": 16442 }, { "epoch": 5.1, "learning_rate": 3.4262852570514106e-06, "loss": 0.0054, "step": 16443 }, { "epoch": 5.1, "learning_rate": 3.4258851770354073e-06, "loss": 0.0219, "step": 16444 }, { "epoch": 5.1, "learning_rate": 3.4254850970194044e-06, "loss": 0.0072, "step": 16445 }, { "epoch": 5.1, "learning_rate": 3.425085017003401e-06, "loss": 0.0036, "step": 16446 }, { "epoch": 5.1, "learning_rate": 3.424684936987398e-06, "loss": 0.0072, "step": 16447 }, { "epoch": 5.11, "learning_rate": 3.4242848569713944e-06, "loss": 0.0096, "step": 16448 }, { "epoch": 5.11, "learning_rate": 3.423884776955391e-06, "loss": 0.0033, "step": 16449 }, { "epoch": 5.11, "learning_rate": 3.4234846969393882e-06, "loss": 0.081, "step": 16450 }, { "epoch": 5.11, "learning_rate": 3.423084616923385e-06, "loss": 0.0021, "step": 16451 }, { "epoch": 5.11, "learning_rate": 3.422684536907382e-06, "loss": 0.0078, "step": 16452 }, { "epoch": 5.11, "learning_rate": 3.4222844568913787e-06, "loss": 0.0172, "step": 16453 }, { "epoch": 5.11, "learning_rate": 3.421884376875375e-06, "loss": 0.0028, "step": 16454 }, { "epoch": 5.11, "learning_rate": 3.421484296859372e-06, "loss": 0.0138, "step": 16455 }, { "epoch": 5.11, "learning_rate": 3.4210842168433688e-06, "loss": 0.0125, "step": 16456 }, { "epoch": 5.11, "learning_rate": 3.420684136827366e-06, "loss": 0.0109, "step": 16457 }, { "epoch": 5.11, "learning_rate": 3.4202840568113626e-06, "loss": 0.0019, "step": 16458 }, { "epoch": 5.11, "learning_rate": 3.4198839767953597e-06, "loss": 0.0124, "step": 16459 }, { "epoch": 5.11, "learning_rate": 3.419483896779356e-06, "loss": 0.0025, "step": 16460 }, { "epoch": 5.11, "learning_rate": 3.4190838167633526e-06, "loss": 0.0043, "step": 16461 }, { "epoch": 5.11, "learning_rate": 3.4186837367473497e-06, "loss": 0.0138, "step": 16462 }, { "epoch": 5.11, "learning_rate": 3.4182836567313464e-06, "loss": 0.0087, "step": 16463 }, { "epoch": 5.11, "learning_rate": 3.4178835767153435e-06, "loss": 0.0159, "step": 16464 }, { "epoch": 5.11, "learning_rate": 3.41748349669934e-06, "loss": 0.0124, "step": 16465 }, { "epoch": 5.11, "learning_rate": 3.4170834166833373e-06, "loss": 0.0143, "step": 16466 }, { "epoch": 5.11, "learning_rate": 3.4166833366673336e-06, "loss": 0.008, "step": 16467 }, { "epoch": 5.11, "learning_rate": 3.4162832566513303e-06, "loss": 0.0372, "step": 16468 }, { "epoch": 5.11, "learning_rate": 3.4158831766353274e-06, "loss": 0.0027, "step": 16469 }, { "epoch": 5.11, "learning_rate": 3.415483096619324e-06, "loss": 0.0117, "step": 16470 }, { "epoch": 5.11, "learning_rate": 3.415083016603321e-06, "loss": 0.0096, "step": 16471 }, { "epoch": 5.11, "learning_rate": 3.414682936587318e-06, "loss": 0.0025, "step": 16472 }, { "epoch": 5.11, "learning_rate": 3.414282856571314e-06, "loss": 0.0037, "step": 16473 }, { "epoch": 5.11, "learning_rate": 3.4138827765553112e-06, "loss": 0.0043, "step": 16474 }, { "epoch": 5.11, "learning_rate": 3.413482696539308e-06, "loss": 0.0034, "step": 16475 }, { "epoch": 5.11, "learning_rate": 3.413082616523305e-06, "loss": 0.007, "step": 16476 }, { "epoch": 5.11, "learning_rate": 3.4126825365073017e-06, "loss": 0.0083, "step": 16477 }, { "epoch": 5.11, "learning_rate": 3.412282456491299e-06, "loss": 0.0282, "step": 16478 }, { "epoch": 5.11, "learning_rate": 3.411882376475295e-06, "loss": 0.0032, "step": 16479 }, { "epoch": 5.12, "learning_rate": 3.4114822964592918e-06, "loss": 0.0342, "step": 16480 }, { "epoch": 5.12, "learning_rate": 3.411082216443289e-06, "loss": 0.0045, "step": 16481 }, { "epoch": 5.12, "learning_rate": 3.4106821364272856e-06, "loss": 0.0308, "step": 16482 }, { "epoch": 5.12, "learning_rate": 3.4102820564112827e-06, "loss": 0.0107, "step": 16483 }, { "epoch": 5.12, "learning_rate": 3.4098819763952794e-06, "loss": 0.0139, "step": 16484 }, { "epoch": 5.12, "learning_rate": 3.4094818963792765e-06, "loss": 0.0034, "step": 16485 }, { "epoch": 5.12, "learning_rate": 3.4090818163632727e-06, "loss": 0.0027, "step": 16486 }, { "epoch": 5.12, "learning_rate": 3.4086817363472694e-06, "loss": 0.0259, "step": 16487 }, { "epoch": 5.12, "learning_rate": 3.4082816563312665e-06, "loss": 0.0079, "step": 16488 }, { "epoch": 5.12, "learning_rate": 3.4078815763152632e-06, "loss": 0.0052, "step": 16489 }, { "epoch": 5.12, "learning_rate": 3.4074814962992603e-06, "loss": 0.0267, "step": 16490 }, { "epoch": 5.12, "learning_rate": 3.407081416283257e-06, "loss": 0.0044, "step": 16491 }, { "epoch": 5.12, "learning_rate": 3.4066813362672537e-06, "loss": 0.0069, "step": 16492 }, { "epoch": 5.12, "learning_rate": 3.4062812562512504e-06, "loss": 0.0026, "step": 16493 }, { "epoch": 5.12, "learning_rate": 3.405881176235247e-06, "loss": 0.0035, "step": 16494 }, { "epoch": 5.12, "learning_rate": 3.405481096219244e-06, "loss": 0.0046, "step": 16495 }, { "epoch": 5.12, "learning_rate": 3.405081016203241e-06, "loss": 0.0036, "step": 16496 }, { "epoch": 5.12, "learning_rate": 3.404680936187238e-06, "loss": 0.0097, "step": 16497 }, { "epoch": 5.12, "learning_rate": 3.4042808561712342e-06, "loss": 0.0096, "step": 16498 }, { "epoch": 5.12, "learning_rate": 3.4038807761552314e-06, "loss": 0.007, "step": 16499 }, { "epoch": 5.12, "learning_rate": 3.403480696139228e-06, "loss": 0.0056, "step": 16500 }, { "epoch": 5.12, "learning_rate": 3.4030806161232247e-06, "loss": 0.0031, "step": 16501 }, { "epoch": 5.12, "learning_rate": 3.402680536107222e-06, "loss": 0.0067, "step": 16502 }, { "epoch": 5.12, "learning_rate": 3.4022804560912185e-06, "loss": 0.004, "step": 16503 }, { "epoch": 5.12, "learning_rate": 3.4018803760752156e-06, "loss": 0.0026, "step": 16504 }, { "epoch": 5.12, "learning_rate": 3.401480296059212e-06, "loss": 0.0261, "step": 16505 }, { "epoch": 5.12, "learning_rate": 3.401080216043209e-06, "loss": 0.0026, "step": 16506 }, { "epoch": 5.12, "learning_rate": 3.4006801360272057e-06, "loss": 0.0067, "step": 16507 }, { "epoch": 5.12, "learning_rate": 3.400280056011203e-06, "loss": 0.1016, "step": 16508 }, { "epoch": 5.12, "learning_rate": 3.3998799759951995e-06, "loss": 0.0063, "step": 16509 }, { "epoch": 5.12, "learning_rate": 3.399479895979196e-06, "loss": 0.0086, "step": 16510 }, { "epoch": 5.12, "learning_rate": 3.399079815963193e-06, "loss": 0.0041, "step": 16511 }, { "epoch": 5.12, "learning_rate": 3.3986797359471895e-06, "loss": 0.0018, "step": 16512 }, { "epoch": 5.13, "learning_rate": 3.3982796559311867e-06, "loss": 0.0236, "step": 16513 }, { "epoch": 5.13, "learning_rate": 3.3978795759151833e-06, "loss": 0.113, "step": 16514 }, { "epoch": 5.13, "learning_rate": 3.3974794958991805e-06, "loss": 0.0033, "step": 16515 }, { "epoch": 5.13, "learning_rate": 3.397079415883177e-06, "loss": 0.005, "step": 16516 }, { "epoch": 5.13, "learning_rate": 3.3966793358671734e-06, "loss": 0.0026, "step": 16517 }, { "epoch": 5.13, "learning_rate": 3.3962792558511705e-06, "loss": 0.0229, "step": 16518 }, { "epoch": 5.13, "learning_rate": 3.395879175835167e-06, "loss": 0.0056, "step": 16519 }, { "epoch": 5.13, "learning_rate": 3.3954790958191643e-06, "loss": 0.0041, "step": 16520 }, { "epoch": 5.13, "learning_rate": 3.395079015803161e-06, "loss": 0.0186, "step": 16521 }, { "epoch": 5.13, "learning_rate": 3.394678935787158e-06, "loss": 0.0585, "step": 16522 }, { "epoch": 5.13, "learning_rate": 3.3942788557711544e-06, "loss": 0.0161, "step": 16523 }, { "epoch": 5.13, "learning_rate": 3.393878775755151e-06, "loss": 0.0023, "step": 16524 }, { "epoch": 5.13, "learning_rate": 3.393478695739148e-06, "loss": 0.0066, "step": 16525 }, { "epoch": 5.13, "learning_rate": 3.393078615723145e-06, "loss": 0.0048, "step": 16526 }, { "epoch": 5.13, "learning_rate": 3.392678535707142e-06, "loss": 0.0043, "step": 16527 }, { "epoch": 5.13, "learning_rate": 3.3922784556911386e-06, "loss": 0.0033, "step": 16528 }, { "epoch": 5.13, "learning_rate": 3.3918783756751358e-06, "loss": 0.0119, "step": 16529 }, { "epoch": 5.13, "learning_rate": 3.391478295659132e-06, "loss": 0.0025, "step": 16530 }, { "epoch": 5.13, "learning_rate": 3.3910782156431287e-06, "loss": 0.0056, "step": 16531 }, { "epoch": 5.13, "learning_rate": 3.390678135627126e-06, "loss": 0.0477, "step": 16532 }, { "epoch": 5.13, "learning_rate": 3.3902780556111225e-06, "loss": 0.0192, "step": 16533 }, { "epoch": 5.13, "learning_rate": 3.3898779755951196e-06, "loss": 0.0023, "step": 16534 }, { "epoch": 5.13, "learning_rate": 3.3894778955791163e-06, "loss": 0.078, "step": 16535 }, { "epoch": 5.13, "learning_rate": 3.3890778155631126e-06, "loss": 0.0148, "step": 16536 }, { "epoch": 5.13, "learning_rate": 3.3886777355471097e-06, "loss": 0.0225, "step": 16537 }, { "epoch": 5.13, "learning_rate": 3.3882776555311064e-06, "loss": 0.0573, "step": 16538 }, { "epoch": 5.13, "learning_rate": 3.3878775755151035e-06, "loss": 0.0076, "step": 16539 }, { "epoch": 5.13, "learning_rate": 3.3874774954991e-06, "loss": 0.0061, "step": 16540 }, { "epoch": 5.13, "learning_rate": 3.3870774154830973e-06, "loss": 0.0175, "step": 16541 }, { "epoch": 5.13, "learning_rate": 3.3866773354670935e-06, "loss": 0.0129, "step": 16542 }, { "epoch": 5.13, "learning_rate": 3.38627725545109e-06, "loss": 0.0064, "step": 16543 }, { "epoch": 5.13, "learning_rate": 3.3858771754350873e-06, "loss": 0.004, "step": 16544 }, { "epoch": 5.14, "learning_rate": 3.385477095419084e-06, "loss": 0.008, "step": 16545 }, { "epoch": 5.14, "learning_rate": 3.385077015403081e-06, "loss": 0.0046, "step": 16546 }, { "epoch": 5.14, "learning_rate": 3.384676935387078e-06, "loss": 0.0134, "step": 16547 }, { "epoch": 5.14, "learning_rate": 3.384276855371075e-06, "loss": 0.0033, "step": 16548 }, { "epoch": 5.14, "learning_rate": 3.383876775355071e-06, "loss": 0.0102, "step": 16549 }, { "epoch": 5.14, "learning_rate": 3.383476695339068e-06, "loss": 0.0076, "step": 16550 }, { "epoch": 5.14, "learning_rate": 3.383076615323065e-06, "loss": 0.0073, "step": 16551 }, { "epoch": 5.14, "learning_rate": 3.3826765353070617e-06, "loss": 0.0071, "step": 16552 }, { "epoch": 5.14, "learning_rate": 3.3822764552910588e-06, "loss": 0.0035, "step": 16553 }, { "epoch": 5.14, "learning_rate": 3.3818763752750555e-06, "loss": 0.0089, "step": 16554 }, { "epoch": 5.14, "learning_rate": 3.3814762952590517e-06, "loss": 0.0062, "step": 16555 }, { "epoch": 5.14, "learning_rate": 3.381076215243049e-06, "loss": 0.0066, "step": 16556 }, { "epoch": 5.14, "learning_rate": 3.3806761352270455e-06, "loss": 0.1164, "step": 16557 }, { "epoch": 5.14, "learning_rate": 3.3802760552110426e-06, "loss": 0.0646, "step": 16558 }, { "epoch": 5.14, "learning_rate": 3.3798759751950393e-06, "loss": 0.0142, "step": 16559 }, { "epoch": 5.14, "learning_rate": 3.3794758951790364e-06, "loss": 0.0038, "step": 16560 }, { "epoch": 5.14, "learning_rate": 3.3790758151630327e-06, "loss": 0.0125, "step": 16561 }, { "epoch": 5.14, "learning_rate": 3.3786757351470294e-06, "loss": 0.0044, "step": 16562 }, { "epoch": 5.14, "learning_rate": 3.3782756551310265e-06, "loss": 0.0037, "step": 16563 }, { "epoch": 5.14, "learning_rate": 3.377875575115023e-06, "loss": 0.0047, "step": 16564 }, { "epoch": 5.14, "learning_rate": 3.3774754950990203e-06, "loss": 0.0038, "step": 16565 }, { "epoch": 5.14, "learning_rate": 3.377075415083017e-06, "loss": 0.0073, "step": 16566 }, { "epoch": 5.14, "learning_rate": 3.376675335067014e-06, "loss": 0.009, "step": 16567 }, { "epoch": 5.14, "learning_rate": 3.3762752550510103e-06, "loss": 0.0092, "step": 16568 }, { "epoch": 5.14, "learning_rate": 3.375875175035007e-06, "loss": 0.0026, "step": 16569 }, { "epoch": 5.14, "learning_rate": 3.375475095019004e-06, "loss": 0.0066, "step": 16570 }, { "epoch": 5.14, "learning_rate": 3.375075015003001e-06, "loss": 0.0028, "step": 16571 }, { "epoch": 5.14, "learning_rate": 3.374674934986998e-06, "loss": 0.0089, "step": 16572 }, { "epoch": 5.14, "learning_rate": 3.3742748549709946e-06, "loss": 0.0233, "step": 16573 }, { "epoch": 5.14, "learning_rate": 3.373874774954991e-06, "loss": 0.0047, "step": 16574 }, { "epoch": 5.14, "learning_rate": 3.373474694938988e-06, "loss": 0.0458, "step": 16575 }, { "epoch": 5.14, "learning_rate": 3.3730746149229847e-06, "loss": 0.0036, "step": 16576 }, { "epoch": 5.15, "learning_rate": 3.3726745349069818e-06, "loss": 0.0093, "step": 16577 }, { "epoch": 5.15, "learning_rate": 3.3722744548909785e-06, "loss": 0.0024, "step": 16578 }, { "epoch": 5.15, "learning_rate": 3.3718743748749756e-06, "loss": 0.0079, "step": 16579 }, { "epoch": 5.15, "learning_rate": 3.371474294858972e-06, "loss": 0.0041, "step": 16580 }, { "epoch": 5.15, "learning_rate": 3.3710742148429685e-06, "loss": 0.0082, "step": 16581 }, { "epoch": 5.15, "learning_rate": 3.3706741348269656e-06, "loss": 0.01, "step": 16582 }, { "epoch": 5.15, "learning_rate": 3.3702740548109623e-06, "loss": 0.002, "step": 16583 }, { "epoch": 5.15, "learning_rate": 3.3698739747949594e-06, "loss": 0.0032, "step": 16584 }, { "epoch": 5.15, "learning_rate": 3.369473894778956e-06, "loss": 0.0055, "step": 16585 }, { "epoch": 5.15, "learning_rate": 3.3690738147629532e-06, "loss": 0.0011, "step": 16586 }, { "epoch": 5.15, "learning_rate": 3.3686737347469495e-06, "loss": 0.0038, "step": 16587 }, { "epoch": 5.15, "learning_rate": 3.368273654730946e-06, "loss": 0.0034, "step": 16588 }, { "epoch": 5.15, "learning_rate": 3.3678735747149433e-06, "loss": 0.0047, "step": 16589 }, { "epoch": 5.15, "learning_rate": 3.36747349469894e-06, "loss": 0.0017, "step": 16590 }, { "epoch": 5.15, "learning_rate": 3.367073414682937e-06, "loss": 0.0055, "step": 16591 }, { "epoch": 5.15, "learning_rate": 3.3666733346669338e-06, "loss": 0.0089, "step": 16592 }, { "epoch": 5.15, "learning_rate": 3.3662732546509305e-06, "loss": 0.0114, "step": 16593 }, { "epoch": 5.15, "learning_rate": 3.365873174634927e-06, "loss": 0.0055, "step": 16594 }, { "epoch": 5.15, "learning_rate": 3.365473094618924e-06, "loss": 0.0185, "step": 16595 }, { "epoch": 5.15, "learning_rate": 3.365073014602921e-06, "loss": 0.0184, "step": 16596 }, { "epoch": 5.15, "learning_rate": 3.3646729345869176e-06, "loss": 0.0197, "step": 16597 }, { "epoch": 5.15, "learning_rate": 3.3642728545709147e-06, "loss": 0.0056, "step": 16598 }, { "epoch": 5.15, "learning_rate": 3.363872774554911e-06, "loss": 0.0288, "step": 16599 }, { "epoch": 5.15, "learning_rate": 3.363472694538908e-06, "loss": 0.0142, "step": 16600 }, { "epoch": 5.15, "learning_rate": 3.363072614522905e-06, "loss": 0.0087, "step": 16601 }, { "epoch": 5.15, "learning_rate": 3.3626725345069015e-06, "loss": 0.0068, "step": 16602 }, { "epoch": 5.15, "learning_rate": 3.3622724544908986e-06, "loss": 0.0092, "step": 16603 }, { "epoch": 5.15, "learning_rate": 3.3618723744748953e-06, "loss": 0.0139, "step": 16604 }, { "epoch": 5.15, "learning_rate": 3.361472294458892e-06, "loss": 0.0118, "step": 16605 }, { "epoch": 5.15, "learning_rate": 3.3610722144428886e-06, "loss": 0.0067, "step": 16606 }, { "epoch": 5.15, "learning_rate": 3.3606721344268858e-06, "loss": 0.0077, "step": 16607 }, { "epoch": 5.15, "learning_rate": 3.3602720544108824e-06, "loss": 0.0013, "step": 16608 }, { "epoch": 5.16, "learning_rate": 3.359871974394879e-06, "loss": 0.007, "step": 16609 }, { "epoch": 5.16, "learning_rate": 3.3594718943788762e-06, "loss": 0.0038, "step": 16610 }, { "epoch": 5.16, "learning_rate": 3.359071814362873e-06, "loss": 0.0105, "step": 16611 }, { "epoch": 5.16, "learning_rate": 3.3586717343468696e-06, "loss": 0.0045, "step": 16612 }, { "epoch": 5.16, "learning_rate": 3.3582716543308663e-06, "loss": 0.0035, "step": 16613 }, { "epoch": 5.16, "learning_rate": 3.3578715743148634e-06, "loss": 0.0048, "step": 16614 }, { "epoch": 5.16, "learning_rate": 3.35747149429886e-06, "loss": 0.0067, "step": 16615 }, { "epoch": 5.16, "learning_rate": 3.3570714142828568e-06, "loss": 0.0034, "step": 16616 }, { "epoch": 5.16, "learning_rate": 3.356671334266854e-06, "loss": 0.0095, "step": 16617 }, { "epoch": 5.16, "learning_rate": 3.35627125425085e-06, "loss": 0.0029, "step": 16618 }, { "epoch": 5.16, "learning_rate": 3.3558711742348473e-06, "loss": 0.005, "step": 16619 }, { "epoch": 5.16, "learning_rate": 3.355471094218844e-06, "loss": 0.0365, "step": 16620 }, { "epoch": 5.16, "learning_rate": 3.355071014202841e-06, "loss": 0.0033, "step": 16621 }, { "epoch": 5.16, "learning_rate": 3.3546709341868377e-06, "loss": 0.0029, "step": 16622 }, { "epoch": 5.16, "learning_rate": 3.354270854170835e-06, "loss": 0.0024, "step": 16623 }, { "epoch": 5.16, "learning_rate": 3.353870774154831e-06, "loss": 0.0059, "step": 16624 }, { "epoch": 5.16, "learning_rate": 3.353470694138828e-06, "loss": 0.0063, "step": 16625 }, { "epoch": 5.16, "learning_rate": 3.353070614122825e-06, "loss": 0.0042, "step": 16626 }, { "epoch": 5.16, "learning_rate": 3.3526705341068216e-06, "loss": 0.0061, "step": 16627 }, { "epoch": 5.16, "learning_rate": 3.3522704540908187e-06, "loss": 0.0027, "step": 16628 }, { "epoch": 5.16, "learning_rate": 3.3518703740748154e-06, "loss": 0.0041, "step": 16629 }, { "epoch": 5.16, "learning_rate": 3.3514702940588125e-06, "loss": 0.0053, "step": 16630 }, { "epoch": 5.16, "learning_rate": 3.3510702140428088e-06, "loss": 0.0085, "step": 16631 }, { "epoch": 5.16, "learning_rate": 3.3506701340268055e-06, "loss": 0.0039, "step": 16632 }, { "epoch": 5.16, "learning_rate": 3.3502700540108026e-06, "loss": 0.0063, "step": 16633 }, { "epoch": 5.16, "learning_rate": 3.3498699739947992e-06, "loss": 0.0063, "step": 16634 }, { "epoch": 5.16, "learning_rate": 3.3494698939787964e-06, "loss": 0.0057, "step": 16635 }, { "epoch": 5.16, "learning_rate": 3.349069813962793e-06, "loss": 0.0042, "step": 16636 }, { "epoch": 5.16, "learning_rate": 3.3486697339467893e-06, "loss": 0.0057, "step": 16637 }, { "epoch": 5.16, "learning_rate": 3.3482696539307864e-06, "loss": 0.0041, "step": 16638 }, { "epoch": 5.16, "learning_rate": 3.347869573914783e-06, "loss": 0.0061, "step": 16639 }, { "epoch": 5.16, "learning_rate": 3.3474694938987802e-06, "loss": 0.002, "step": 16640 }, { "epoch": 5.17, "learning_rate": 3.347069413882777e-06, "loss": 0.0218, "step": 16641 }, { "epoch": 5.17, "learning_rate": 3.346669333866774e-06, "loss": 0.0041, "step": 16642 }, { "epoch": 5.17, "learning_rate": 3.3462692538507703e-06, "loss": 0.006, "step": 16643 }, { "epoch": 5.17, "learning_rate": 3.345869173834767e-06, "loss": 0.0058, "step": 16644 }, { "epoch": 5.17, "learning_rate": 3.345469093818764e-06, "loss": 0.0087, "step": 16645 }, { "epoch": 5.17, "learning_rate": 3.3450690138027608e-06, "loss": 0.0029, "step": 16646 }, { "epoch": 5.17, "learning_rate": 3.344668933786758e-06, "loss": 0.0026, "step": 16647 }, { "epoch": 5.17, "learning_rate": 3.3442688537707545e-06, "loss": 0.0067, "step": 16648 }, { "epoch": 5.17, "learning_rate": 3.3438687737547517e-06, "loss": 0.0055, "step": 16649 }, { "epoch": 5.17, "learning_rate": 3.343468693738748e-06, "loss": 0.0044, "step": 16650 }, { "epoch": 5.17, "learning_rate": 3.3430686137227446e-06, "loss": 0.0084, "step": 16651 }, { "epoch": 5.17, "learning_rate": 3.3426685337067417e-06, "loss": 0.0141, "step": 16652 }, { "epoch": 5.17, "learning_rate": 3.3422684536907384e-06, "loss": 0.0037, "step": 16653 }, { "epoch": 5.17, "learning_rate": 3.3418683736747355e-06, "loss": 0.0073, "step": 16654 }, { "epoch": 5.17, "learning_rate": 3.341468293658732e-06, "loss": 0.0128, "step": 16655 }, { "epoch": 5.17, "learning_rate": 3.3410682136427285e-06, "loss": 0.0031, "step": 16656 }, { "epoch": 5.17, "learning_rate": 3.3406681336267256e-06, "loss": 0.0184, "step": 16657 }, { "epoch": 5.17, "learning_rate": 3.3402680536107223e-06, "loss": 0.002, "step": 16658 }, { "epoch": 5.17, "learning_rate": 3.3398679735947194e-06, "loss": 0.0107, "step": 16659 }, { "epoch": 5.17, "learning_rate": 3.339467893578716e-06, "loss": 0.0132, "step": 16660 }, { "epoch": 5.17, "learning_rate": 3.339067813562713e-06, "loss": 0.0078, "step": 16661 }, { "epoch": 5.17, "learning_rate": 3.3386677335467094e-06, "loss": 0.0057, "step": 16662 }, { "epoch": 5.17, "learning_rate": 3.338267653530706e-06, "loss": 0.0106, "step": 16663 }, { "epoch": 5.17, "learning_rate": 3.3378675735147032e-06, "loss": 0.0023, "step": 16664 }, { "epoch": 5.17, "learning_rate": 3.3374674934987e-06, "loss": 0.0018, "step": 16665 }, { "epoch": 5.17, "learning_rate": 3.337067413482697e-06, "loss": 0.0032, "step": 16666 }, { "epoch": 5.17, "learning_rate": 3.3366673334666937e-06, "loss": 0.0072, "step": 16667 }, { "epoch": 5.17, "learning_rate": 3.33626725345069e-06, "loss": 0.0128, "step": 16668 }, { "epoch": 5.17, "learning_rate": 3.335867173434687e-06, "loss": 0.0029, "step": 16669 }, { "epoch": 5.17, "learning_rate": 3.3354670934186838e-06, "loss": 0.0065, "step": 16670 }, { "epoch": 5.17, "learning_rate": 3.335067013402681e-06, "loss": 0.0092, "step": 16671 }, { "epoch": 5.17, "learning_rate": 3.3346669333866776e-06, "loss": 0.0019, "step": 16672 }, { "epoch": 5.17, "learning_rate": 3.3342668533706747e-06, "loss": 0.0103, "step": 16673 }, { "epoch": 5.18, "learning_rate": 3.3338667733546714e-06, "loss": 0.0082, "step": 16674 }, { "epoch": 5.18, "learning_rate": 3.3334666933386676e-06, "loss": 0.009, "step": 16675 }, { "epoch": 5.18, "learning_rate": 3.3330666133226647e-06, "loss": 0.0044, "step": 16676 }, { "epoch": 5.18, "learning_rate": 3.3326665333066614e-06, "loss": 0.0277, "step": 16677 }, { "epoch": 5.18, "learning_rate": 3.3322664532906585e-06, "loss": 0.0156, "step": 16678 }, { "epoch": 5.18, "learning_rate": 3.3318663732746552e-06, "loss": 0.0055, "step": 16679 }, { "epoch": 5.18, "learning_rate": 3.3314662932586523e-06, "loss": 0.0449, "step": 16680 }, { "epoch": 5.18, "learning_rate": 3.3310662132426486e-06, "loss": 0.0086, "step": 16681 }, { "epoch": 5.18, "learning_rate": 3.3306661332266453e-06, "loss": 0.002, "step": 16682 }, { "epoch": 5.18, "learning_rate": 3.3302660532106424e-06, "loss": 0.0045, "step": 16683 }, { "epoch": 5.18, "learning_rate": 3.329865973194639e-06, "loss": 0.0109, "step": 16684 }, { "epoch": 5.18, "learning_rate": 3.329465893178636e-06, "loss": 0.0139, "step": 16685 }, { "epoch": 5.18, "learning_rate": 3.329065813162633e-06, "loss": 0.0114, "step": 16686 }, { "epoch": 5.18, "learning_rate": 3.328665733146629e-06, "loss": 0.0114, "step": 16687 }, { "epoch": 5.18, "learning_rate": 3.3282656531306262e-06, "loss": 0.0033, "step": 16688 }, { "epoch": 5.18, "learning_rate": 3.327865573114623e-06, "loss": 0.037, "step": 16689 }, { "epoch": 5.18, "learning_rate": 3.32746549309862e-06, "loss": 0.0046, "step": 16690 }, { "epoch": 5.18, "learning_rate": 3.3270654130826167e-06, "loss": 0.0153, "step": 16691 }, { "epoch": 5.18, "learning_rate": 3.326665333066614e-06, "loss": 0.0128, "step": 16692 }, { "epoch": 5.18, "learning_rate": 3.3262652530506105e-06, "loss": 0.0078, "step": 16693 }, { "epoch": 5.18, "learning_rate": 3.3258651730346068e-06, "loss": 0.0068, "step": 16694 }, { "epoch": 5.18, "learning_rate": 3.325465093018604e-06, "loss": 0.0415, "step": 16695 }, { "epoch": 5.18, "learning_rate": 3.3250650130026006e-06, "loss": 0.0023, "step": 16696 }, { "epoch": 5.18, "learning_rate": 3.3246649329865977e-06, "loss": 0.0044, "step": 16697 }, { "epoch": 5.18, "learning_rate": 3.3242648529705944e-06, "loss": 0.0072, "step": 16698 }, { "epoch": 5.18, "learning_rate": 3.3238647729545915e-06, "loss": 0.0144, "step": 16699 }, { "epoch": 5.18, "learning_rate": 3.3234646929385877e-06, "loss": 0.0116, "step": 16700 }, { "epoch": 5.18, "learning_rate": 3.323064612922585e-06, "loss": 0.0106, "step": 16701 }, { "epoch": 5.18, "learning_rate": 3.3226645329065815e-06, "loss": 0.0044, "step": 16702 }, { "epoch": 5.18, "learning_rate": 3.3222644528905782e-06, "loss": 0.0135, "step": 16703 }, { "epoch": 5.18, "learning_rate": 3.3218643728745753e-06, "loss": 0.0046, "step": 16704 }, { "epoch": 5.18, "learning_rate": 3.321464292858572e-06, "loss": 0.0106, "step": 16705 }, { "epoch": 5.19, "learning_rate": 3.3210642128425687e-06, "loss": 0.0018, "step": 16706 }, { "epoch": 5.19, "learning_rate": 3.3206641328265654e-06, "loss": 0.0026, "step": 16707 }, { "epoch": 5.19, "learning_rate": 3.3202640528105625e-06, "loss": 0.0141, "step": 16708 }, { "epoch": 5.19, "learning_rate": 3.319863972794559e-06, "loss": 0.0046, "step": 16709 }, { "epoch": 5.19, "learning_rate": 3.319463892778556e-06, "loss": 0.0059, "step": 16710 }, { "epoch": 5.19, "learning_rate": 3.319063812762553e-06, "loss": 0.002, "step": 16711 }, { "epoch": 5.19, "learning_rate": 3.3186637327465497e-06, "loss": 0.0166, "step": 16712 }, { "epoch": 5.19, "learning_rate": 3.3182636527305464e-06, "loss": 0.0072, "step": 16713 }, { "epoch": 5.19, "learning_rate": 3.317863572714543e-06, "loss": 0.0122, "step": 16714 }, { "epoch": 5.19, "learning_rate": 3.31746349269854e-06, "loss": 0.0071, "step": 16715 }, { "epoch": 5.19, "learning_rate": 3.317063412682537e-06, "loss": 0.0039, "step": 16716 }, { "epoch": 5.19, "learning_rate": 3.3166633326665335e-06, "loss": 0.0083, "step": 16717 }, { "epoch": 5.19, "learning_rate": 3.3162632526505306e-06, "loss": 0.0162, "step": 16718 }, { "epoch": 5.19, "learning_rate": 3.315863172634527e-06, "loss": 0.0124, "step": 16719 }, { "epoch": 5.19, "learning_rate": 3.315463092618524e-06, "loss": 0.0068, "step": 16720 }, { "epoch": 5.19, "learning_rate": 3.3150630126025207e-06, "loss": 0.003, "step": 16721 }, { "epoch": 5.19, "learning_rate": 3.314662932586518e-06, "loss": 0.016, "step": 16722 }, { "epoch": 5.19, "learning_rate": 3.3142628525705145e-06, "loss": 0.0063, "step": 16723 }, { "epoch": 5.19, "learning_rate": 3.313862772554511e-06, "loss": 0.019, "step": 16724 }, { "epoch": 5.19, "learning_rate": 3.313462692538508e-06, "loss": 0.0013, "step": 16725 }, { "epoch": 5.19, "learning_rate": 3.3130626125225045e-06, "loss": 0.0076, "step": 16726 }, { "epoch": 5.19, "learning_rate": 3.3126625325065017e-06, "loss": 0.043, "step": 16727 }, { "epoch": 5.19, "learning_rate": 3.3122624524904983e-06, "loss": 0.0149, "step": 16728 }, { "epoch": 5.19, "learning_rate": 3.3118623724744955e-06, "loss": 0.0795, "step": 16729 }, { "epoch": 5.19, "learning_rate": 3.311462292458492e-06, "loss": 0.0218, "step": 16730 }, { "epoch": 5.19, "learning_rate": 3.3110622124424893e-06, "loss": 0.0048, "step": 16731 }, { "epoch": 5.19, "learning_rate": 3.3106621324264855e-06, "loss": 0.0108, "step": 16732 }, { "epoch": 5.19, "learning_rate": 3.310262052410482e-06, "loss": 0.0039, "step": 16733 }, { "epoch": 5.19, "learning_rate": 3.3098619723944793e-06, "loss": 0.0109, "step": 16734 }, { "epoch": 5.19, "learning_rate": 3.309461892378476e-06, "loss": 0.0108, "step": 16735 }, { "epoch": 5.19, "learning_rate": 3.309061812362473e-06, "loss": 0.0045, "step": 16736 }, { "epoch": 5.19, "learning_rate": 3.30866173234647e-06, "loss": 0.003, "step": 16737 }, { "epoch": 5.2, "learning_rate": 3.308261652330466e-06, "loss": 0.0067, "step": 16738 }, { "epoch": 5.2, "learning_rate": 3.307861572314463e-06, "loss": 0.0065, "step": 16739 }, { "epoch": 5.2, "learning_rate": 3.30746149229846e-06, "loss": 0.0063, "step": 16740 }, { "epoch": 5.2, "learning_rate": 3.307061412282457e-06, "loss": 0.0091, "step": 16741 }, { "epoch": 5.2, "learning_rate": 3.3066613322664536e-06, "loss": 0.0029, "step": 16742 }, { "epoch": 5.2, "learning_rate": 3.3062612522504508e-06, "loss": 0.0171, "step": 16743 }, { "epoch": 5.2, "learning_rate": 3.305861172234447e-06, "loss": 0.0023, "step": 16744 }, { "epoch": 5.2, "learning_rate": 3.3054610922184437e-06, "loss": 0.0056, "step": 16745 }, { "epoch": 5.2, "learning_rate": 3.305061012202441e-06, "loss": 0.0271, "step": 16746 }, { "epoch": 5.2, "learning_rate": 3.3046609321864375e-06, "loss": 0.0053, "step": 16747 }, { "epoch": 5.2, "learning_rate": 3.3042608521704346e-06, "loss": 0.0062, "step": 16748 }, { "epoch": 5.2, "learning_rate": 3.3038607721544313e-06, "loss": 0.0041, "step": 16749 }, { "epoch": 5.2, "learning_rate": 3.3034606921384276e-06, "loss": 0.1326, "step": 16750 }, { "epoch": 5.2, "learning_rate": 3.3030606121224247e-06, "loss": 0.0071, "step": 16751 }, { "epoch": 5.2, "learning_rate": 3.3026605321064214e-06, "loss": 0.036, "step": 16752 }, { "epoch": 5.2, "learning_rate": 3.3022604520904185e-06, "loss": 0.0048, "step": 16753 }, { "epoch": 5.2, "learning_rate": 3.301860372074415e-06, "loss": 0.02, "step": 16754 }, { "epoch": 5.2, "learning_rate": 3.3014602920584123e-06, "loss": 0.0126, "step": 16755 }, { "epoch": 5.2, "learning_rate": 3.301060212042409e-06, "loss": 0.0022, "step": 16756 }, { "epoch": 5.2, "learning_rate": 3.3006601320264052e-06, "loss": 0.0029, "step": 16757 }, { "epoch": 5.2, "learning_rate": 3.3002600520104023e-06, "loss": 0.0137, "step": 16758 }, { "epoch": 5.2, "learning_rate": 3.299859971994399e-06, "loss": 0.0063, "step": 16759 }, { "epoch": 5.2, "learning_rate": 3.299459891978396e-06, "loss": 0.0067, "step": 16760 }, { "epoch": 5.2, "learning_rate": 3.299059811962393e-06, "loss": 0.0111, "step": 16761 }, { "epoch": 5.2, "learning_rate": 3.29865973194639e-06, "loss": 0.0237, "step": 16762 }, { "epoch": 5.2, "learning_rate": 3.298259651930386e-06, "loss": 0.0049, "step": 16763 }, { "epoch": 5.2, "learning_rate": 3.297859571914383e-06, "loss": 0.0492, "step": 16764 }, { "epoch": 5.2, "learning_rate": 3.29745949189838e-06, "loss": 0.0054, "step": 16765 }, { "epoch": 5.2, "learning_rate": 3.2970594118823767e-06, "loss": 0.0052, "step": 16766 }, { "epoch": 5.2, "learning_rate": 3.2966593318663738e-06, "loss": 0.0029, "step": 16767 }, { "epoch": 5.2, "learning_rate": 3.2962592518503705e-06, "loss": 0.0033, "step": 16768 }, { "epoch": 5.2, "learning_rate": 3.2958591718343667e-06, "loss": 0.0139, "step": 16769 }, { "epoch": 5.21, "learning_rate": 3.295459091818364e-06, "loss": 0.0034, "step": 16770 }, { "epoch": 5.21, "learning_rate": 3.2950590118023605e-06, "loss": 0.027, "step": 16771 }, { "epoch": 5.21, "learning_rate": 3.2946589317863576e-06, "loss": 0.0068, "step": 16772 }, { "epoch": 5.21, "learning_rate": 3.2942588517703543e-06, "loss": 0.0118, "step": 16773 }, { "epoch": 5.21, "learning_rate": 3.2938587717543514e-06, "loss": 0.0023, "step": 16774 }, { "epoch": 5.21, "learning_rate": 3.293458691738348e-06, "loss": 0.0061, "step": 16775 }, { "epoch": 5.21, "learning_rate": 3.2930586117223444e-06, "loss": 0.0093, "step": 16776 }, { "epoch": 5.21, "learning_rate": 3.2926585317063415e-06, "loss": 0.0113, "step": 16777 }, { "epoch": 5.21, "learning_rate": 3.292258451690338e-06, "loss": 0.0058, "step": 16778 }, { "epoch": 5.21, "learning_rate": 3.2918583716743353e-06, "loss": 0.006, "step": 16779 }, { "epoch": 5.21, "learning_rate": 3.291458291658332e-06, "loss": 0.0029, "step": 16780 }, { "epoch": 5.21, "learning_rate": 3.291058211642329e-06, "loss": 0.0059, "step": 16781 }, { "epoch": 5.21, "learning_rate": 3.2906581316263253e-06, "loss": 0.003, "step": 16782 }, { "epoch": 5.21, "learning_rate": 3.290258051610322e-06, "loss": 0.01, "step": 16783 }, { "epoch": 5.21, "learning_rate": 3.289857971594319e-06, "loss": 0.0083, "step": 16784 }, { "epoch": 5.21, "learning_rate": 3.289457891578316e-06, "loss": 0.0101, "step": 16785 }, { "epoch": 5.21, "learning_rate": 3.289057811562313e-06, "loss": 0.0089, "step": 16786 }, { "epoch": 5.21, "learning_rate": 3.2886577315463096e-06, "loss": 0.0066, "step": 16787 }, { "epoch": 5.21, "learning_rate": 3.288257651530306e-06, "loss": 0.0035, "step": 16788 }, { "epoch": 5.21, "learning_rate": 3.287857571514303e-06, "loss": 0.0071, "step": 16789 }, { "epoch": 5.21, "learning_rate": 3.2874574914982997e-06, "loss": 0.002, "step": 16790 }, { "epoch": 5.21, "learning_rate": 3.2870574114822968e-06, "loss": 0.0024, "step": 16791 }, { "epoch": 5.21, "learning_rate": 3.2866573314662935e-06, "loss": 0.0061, "step": 16792 }, { "epoch": 5.21, "learning_rate": 3.2862572514502906e-06, "loss": 0.0058, "step": 16793 }, { "epoch": 5.21, "learning_rate": 3.2858571714342873e-06, "loss": 0.0094, "step": 16794 }, { "epoch": 5.21, "learning_rate": 3.2854570914182835e-06, "loss": 0.0028, "step": 16795 }, { "epoch": 5.21, "learning_rate": 3.2850570114022806e-06, "loss": 0.0035, "step": 16796 }, { "epoch": 5.21, "learning_rate": 3.2846569313862773e-06, "loss": 0.0142, "step": 16797 }, { "epoch": 5.21, "learning_rate": 3.2842568513702744e-06, "loss": 0.0047, "step": 16798 }, { "epoch": 5.21, "learning_rate": 3.283856771354271e-06, "loss": 0.0049, "step": 16799 }, { "epoch": 5.21, "learning_rate": 3.2834566913382682e-06, "loss": 0.0104, "step": 16800 }, { "epoch": 5.21, "learning_rate": 3.2830566113222645e-06, "loss": 0.0065, "step": 16801 }, { "epoch": 5.21, "learning_rate": 3.282656531306261e-06, "loss": 0.0122, "step": 16802 }, { "epoch": 5.22, "learning_rate": 3.2822564512902583e-06, "loss": 0.0086, "step": 16803 }, { "epoch": 5.22, "learning_rate": 3.281856371274255e-06, "loss": 0.0112, "step": 16804 }, { "epoch": 5.22, "learning_rate": 3.281456291258252e-06, "loss": 0.0022, "step": 16805 }, { "epoch": 5.22, "learning_rate": 3.2810562112422488e-06, "loss": 0.0061, "step": 16806 }, { "epoch": 5.22, "learning_rate": 3.2806561312262455e-06, "loss": 0.0585, "step": 16807 }, { "epoch": 5.22, "learning_rate": 3.280256051210242e-06, "loss": 0.003, "step": 16808 }, { "epoch": 5.22, "learning_rate": 3.2798559711942393e-06, "loss": 0.0059, "step": 16809 }, { "epoch": 5.22, "learning_rate": 3.279455891178236e-06, "loss": 0.0061, "step": 16810 }, { "epoch": 5.22, "learning_rate": 3.2790558111622326e-06, "loss": 0.0076, "step": 16811 }, { "epoch": 5.22, "learning_rate": 3.2786557311462297e-06, "loss": 0.0034, "step": 16812 }, { "epoch": 5.22, "learning_rate": 3.2782556511302264e-06, "loss": 0.0055, "step": 16813 }, { "epoch": 5.22, "learning_rate": 3.277855571114223e-06, "loss": 0.0039, "step": 16814 }, { "epoch": 5.22, "learning_rate": 3.27745549109822e-06, "loss": 0.0248, "step": 16815 }, { "epoch": 5.22, "learning_rate": 3.277055411082217e-06, "loss": 0.0045, "step": 16816 }, { "epoch": 5.22, "learning_rate": 3.2766553310662136e-06, "loss": 0.0076, "step": 16817 }, { "epoch": 5.22, "learning_rate": 3.2762552510502103e-06, "loss": 0.0217, "step": 16818 }, { "epoch": 5.22, "learning_rate": 3.2758551710342074e-06, "loss": 0.0061, "step": 16819 }, { "epoch": 5.22, "learning_rate": 3.2754550910182036e-06, "loss": 0.0487, "step": 16820 }, { "epoch": 5.22, "learning_rate": 3.2750550110022008e-06, "loss": 0.003, "step": 16821 }, { "epoch": 5.22, "learning_rate": 3.2746549309861974e-06, "loss": 0.0066, "step": 16822 }, { "epoch": 5.22, "learning_rate": 3.2742548509701946e-06, "loss": 0.005, "step": 16823 }, { "epoch": 5.22, "learning_rate": 3.2738547709541912e-06, "loss": 0.0112, "step": 16824 }, { "epoch": 5.22, "learning_rate": 3.273454690938188e-06, "loss": 0.0028, "step": 16825 }, { "epoch": 5.22, "learning_rate": 3.2730546109221846e-06, "loss": 0.0023, "step": 16826 }, { "epoch": 5.22, "learning_rate": 3.2726545309061813e-06, "loss": 0.0071, "step": 16827 }, { "epoch": 5.22, "learning_rate": 3.2722544508901784e-06, "loss": 0.0342, "step": 16828 }, { "epoch": 5.22, "learning_rate": 3.271854370874175e-06, "loss": 0.0288, "step": 16829 }, { "epoch": 5.22, "learning_rate": 3.271454290858172e-06, "loss": 0.002, "step": 16830 }, { "epoch": 5.22, "learning_rate": 3.271054210842169e-06, "loss": 0.0083, "step": 16831 }, { "epoch": 5.22, "learning_rate": 3.270654130826165e-06, "loss": 0.0055, "step": 16832 }, { "epoch": 5.22, "learning_rate": 3.2702540508101623e-06, "loss": 0.0115, "step": 16833 }, { "epoch": 5.22, "learning_rate": 3.269853970794159e-06, "loss": 0.0031, "step": 16834 }, { "epoch": 5.23, "learning_rate": 3.269453890778156e-06, "loss": 0.0085, "step": 16835 }, { "epoch": 5.23, "learning_rate": 3.2690538107621527e-06, "loss": 0.001, "step": 16836 }, { "epoch": 5.23, "learning_rate": 3.26865373074615e-06, "loss": 0.0075, "step": 16837 }, { "epoch": 5.23, "learning_rate": 3.2682536507301465e-06, "loss": 0.004, "step": 16838 }, { "epoch": 5.23, "learning_rate": 3.267853570714143e-06, "loss": 0.016, "step": 16839 }, { "epoch": 5.23, "learning_rate": 3.26745349069814e-06, "loss": 0.0034, "step": 16840 }, { "epoch": 5.23, "learning_rate": 3.2670534106821366e-06, "loss": 0.0044, "step": 16841 }, { "epoch": 5.23, "learning_rate": 3.2666533306661337e-06, "loss": 0.0058, "step": 16842 }, { "epoch": 5.23, "learning_rate": 3.2662532506501304e-06, "loss": 0.0111, "step": 16843 }, { "epoch": 5.23, "learning_rate": 3.2658531706341275e-06, "loss": 0.0106, "step": 16844 }, { "epoch": 5.23, "learning_rate": 3.2654530906181238e-06, "loss": 0.0078, "step": 16845 }, { "epoch": 5.23, "learning_rate": 3.2650530106021205e-06, "loss": 0.0394, "step": 16846 }, { "epoch": 5.23, "learning_rate": 3.2646529305861176e-06, "loss": 0.0074, "step": 16847 }, { "epoch": 5.23, "learning_rate": 3.2642528505701143e-06, "loss": 0.0034, "step": 16848 }, { "epoch": 5.23, "learning_rate": 3.2638527705541114e-06, "loss": 0.0714, "step": 16849 }, { "epoch": 5.23, "learning_rate": 3.263452690538108e-06, "loss": 0.0329, "step": 16850 }, { "epoch": 5.23, "learning_rate": 3.2630526105221043e-06, "loss": 0.0061, "step": 16851 }, { "epoch": 5.23, "learning_rate": 3.2626525305061014e-06, "loss": 0.0107, "step": 16852 }, { "epoch": 5.23, "learning_rate": 3.262252450490098e-06, "loss": 0.0277, "step": 16853 }, { "epoch": 5.23, "learning_rate": 3.2618523704740952e-06, "loss": 0.0041, "step": 16854 }, { "epoch": 5.23, "learning_rate": 3.261452290458092e-06, "loss": 0.002, "step": 16855 }, { "epoch": 5.23, "learning_rate": 3.261052210442089e-06, "loss": 0.0049, "step": 16856 }, { "epoch": 5.23, "learning_rate": 3.2606521304260857e-06, "loss": 0.0057, "step": 16857 }, { "epoch": 5.23, "learning_rate": 3.260252050410082e-06, "loss": 0.0029, "step": 16858 }, { "epoch": 5.23, "learning_rate": 3.259851970394079e-06, "loss": 0.004, "step": 16859 }, { "epoch": 5.23, "learning_rate": 3.2594518903780758e-06, "loss": 0.0157, "step": 16860 }, { "epoch": 5.23, "learning_rate": 3.259051810362073e-06, "loss": 0.0079, "step": 16861 }, { "epoch": 5.23, "learning_rate": 3.2586517303460696e-06, "loss": 0.0092, "step": 16862 }, { "epoch": 5.23, "learning_rate": 3.2582516503300667e-06, "loss": 0.0335, "step": 16863 }, { "epoch": 5.23, "learning_rate": 3.257851570314063e-06, "loss": 0.0113, "step": 16864 }, { "epoch": 5.23, "learning_rate": 3.2574514902980596e-06, "loss": 0.0092, "step": 16865 }, { "epoch": 5.23, "learning_rate": 3.2570514102820567e-06, "loss": 0.0105, "step": 16866 }, { "epoch": 5.24, "learning_rate": 3.2566513302660534e-06, "loss": 0.0058, "step": 16867 }, { "epoch": 5.24, "learning_rate": 3.2562512502500505e-06, "loss": 0.0121, "step": 16868 }, { "epoch": 5.24, "learning_rate": 3.255851170234047e-06, "loss": 0.0019, "step": 16869 }, { "epoch": 5.24, "learning_rate": 3.2554510902180435e-06, "loss": 0.0112, "step": 16870 }, { "epoch": 5.24, "learning_rate": 3.2550510102020406e-06, "loss": 0.0028, "step": 16871 }, { "epoch": 5.24, "learning_rate": 3.2546509301860373e-06, "loss": 0.003, "step": 16872 }, { "epoch": 5.24, "learning_rate": 3.2542508501700344e-06, "loss": 0.0035, "step": 16873 }, { "epoch": 5.24, "learning_rate": 3.253850770154031e-06, "loss": 0.0138, "step": 16874 }, { "epoch": 5.24, "learning_rate": 3.253450690138028e-06, "loss": 0.0199, "step": 16875 }, { "epoch": 5.24, "learning_rate": 3.253050610122025e-06, "loss": 0.0033, "step": 16876 }, { "epoch": 5.24, "learning_rate": 3.252650530106021e-06, "loss": 0.0105, "step": 16877 }, { "epoch": 5.24, "learning_rate": 3.2522504500900182e-06, "loss": 0.0093, "step": 16878 }, { "epoch": 5.24, "learning_rate": 3.251850370074015e-06, "loss": 0.0178, "step": 16879 }, { "epoch": 5.24, "learning_rate": 3.251450290058012e-06, "loss": 0.0032, "step": 16880 }, { "epoch": 5.24, "learning_rate": 3.2510502100420087e-06, "loss": 0.0148, "step": 16881 }, { "epoch": 5.24, "learning_rate": 3.250650130026006e-06, "loss": 0.0077, "step": 16882 }, { "epoch": 5.24, "learning_rate": 3.250250050010002e-06, "loss": 0.0063, "step": 16883 }, { "epoch": 5.24, "learning_rate": 3.2498499699939988e-06, "loss": 0.0135, "step": 16884 }, { "epoch": 5.24, "learning_rate": 3.249449889977996e-06, "loss": 0.009, "step": 16885 }, { "epoch": 5.24, "learning_rate": 3.2490498099619926e-06, "loss": 0.0065, "step": 16886 }, { "epoch": 5.24, "learning_rate": 3.2486497299459897e-06, "loss": 0.0105, "step": 16887 }, { "epoch": 5.24, "learning_rate": 3.2482496499299864e-06, "loss": 0.0073, "step": 16888 }, { "epoch": 5.24, "learning_rate": 3.2478495699139826e-06, "loss": 0.0091, "step": 16889 }, { "epoch": 5.24, "learning_rate": 3.2474494898979797e-06, "loss": 0.0576, "step": 16890 }, { "epoch": 5.24, "learning_rate": 3.2470494098819764e-06, "loss": 0.0031, "step": 16891 }, { "epoch": 5.24, "learning_rate": 3.2466493298659735e-06, "loss": 0.0108, "step": 16892 }, { "epoch": 5.24, "learning_rate": 3.2462492498499702e-06, "loss": 0.0201, "step": 16893 }, { "epoch": 5.24, "learning_rate": 3.2458491698339673e-06, "loss": 0.005, "step": 16894 }, { "epoch": 5.24, "learning_rate": 3.2454490898179636e-06, "loss": 0.0039, "step": 16895 }, { "epoch": 5.24, "learning_rate": 3.2450490098019603e-06, "loss": 0.007, "step": 16896 }, { "epoch": 5.24, "learning_rate": 3.2446489297859574e-06, "loss": 0.0067, "step": 16897 }, { "epoch": 5.24, "learning_rate": 3.244248849769954e-06, "loss": 0.0029, "step": 16898 }, { "epoch": 5.25, "learning_rate": 3.243848769753951e-06, "loss": 0.0109, "step": 16899 }, { "epoch": 5.25, "learning_rate": 3.243448689737948e-06, "loss": 0.0025, "step": 16900 }, { "epoch": 5.25, "learning_rate": 3.243048609721945e-06, "loss": 0.003, "step": 16901 }, { "epoch": 5.25, "learning_rate": 3.2426485297059412e-06, "loss": 0.006, "step": 16902 }, { "epoch": 5.25, "learning_rate": 3.242248449689938e-06, "loss": 0.0107, "step": 16903 }, { "epoch": 5.25, "learning_rate": 3.241848369673935e-06, "loss": 0.0042, "step": 16904 }, { "epoch": 5.25, "learning_rate": 3.2414482896579317e-06, "loss": 0.0121, "step": 16905 }, { "epoch": 5.25, "learning_rate": 3.241048209641929e-06, "loss": 0.0113, "step": 16906 }, { "epoch": 5.25, "learning_rate": 3.2406481296259255e-06, "loss": 0.0034, "step": 16907 }, { "epoch": 5.25, "learning_rate": 3.240248049609922e-06, "loss": 0.0122, "step": 16908 }, { "epoch": 5.25, "learning_rate": 3.239847969593919e-06, "loss": 0.0077, "step": 16909 }, { "epoch": 5.25, "learning_rate": 3.2394478895779156e-06, "loss": 0.0107, "step": 16910 }, { "epoch": 5.25, "learning_rate": 3.2390478095619127e-06, "loss": 0.0044, "step": 16911 }, { "epoch": 5.25, "learning_rate": 3.2386477295459094e-06, "loss": 0.0228, "step": 16912 }, { "epoch": 5.25, "learning_rate": 3.2382476495299065e-06, "loss": 0.04, "step": 16913 }, { "epoch": 5.25, "learning_rate": 3.2378475695139027e-06, "loss": 0.0375, "step": 16914 }, { "epoch": 5.25, "learning_rate": 3.2374474894979e-06, "loss": 0.0035, "step": 16915 }, { "epoch": 5.25, "learning_rate": 3.2370474094818965e-06, "loss": 0.0041, "step": 16916 }, { "epoch": 5.25, "learning_rate": 3.2366473294658932e-06, "loss": 0.0549, "step": 16917 }, { "epoch": 5.25, "learning_rate": 3.2362472494498903e-06, "loss": 0.0028, "step": 16918 }, { "epoch": 5.25, "learning_rate": 3.235847169433887e-06, "loss": 0.0403, "step": 16919 }, { "epoch": 5.25, "learning_rate": 3.235447089417884e-06, "loss": 0.003, "step": 16920 }, { "epoch": 5.25, "learning_rate": 3.2350470094018804e-06, "loss": 0.0108, "step": 16921 }, { "epoch": 5.25, "learning_rate": 3.2346469293858775e-06, "loss": 0.0095, "step": 16922 }, { "epoch": 5.25, "learning_rate": 3.234246849369874e-06, "loss": 0.0031, "step": 16923 }, { "epoch": 5.25, "learning_rate": 3.2338467693538713e-06, "loss": 0.0376, "step": 16924 }, { "epoch": 5.25, "learning_rate": 3.233446689337868e-06, "loss": 0.0084, "step": 16925 }, { "epoch": 5.25, "learning_rate": 3.2330466093218647e-06, "loss": 0.0075, "step": 16926 }, { "epoch": 5.25, "learning_rate": 3.2326465293058614e-06, "loss": 0.0094, "step": 16927 }, { "epoch": 5.25, "learning_rate": 3.232246449289858e-06, "loss": 0.0117, "step": 16928 }, { "epoch": 5.25, "learning_rate": 3.231846369273855e-06, "loss": 0.0058, "step": 16929 }, { "epoch": 5.25, "learning_rate": 3.231446289257852e-06, "loss": 0.0042, "step": 16930 }, { "epoch": 5.26, "learning_rate": 3.231046209241849e-06, "loss": 0.0067, "step": 16931 }, { "epoch": 5.26, "learning_rate": 3.2306461292258456e-06, "loss": 0.0073, "step": 16932 }, { "epoch": 5.26, "learning_rate": 3.230246049209842e-06, "loss": 0.0083, "step": 16933 }, { "epoch": 5.26, "learning_rate": 3.229845969193839e-06, "loss": 0.0041, "step": 16934 }, { "epoch": 5.26, "learning_rate": 3.2294458891778357e-06, "loss": 0.0314, "step": 16935 }, { "epoch": 5.26, "learning_rate": 3.229045809161833e-06, "loss": 0.0042, "step": 16936 }, { "epoch": 5.26, "learning_rate": 3.2286457291458295e-06, "loss": 0.0053, "step": 16937 }, { "epoch": 5.26, "learning_rate": 3.2282456491298266e-06, "loss": 0.0017, "step": 16938 }, { "epoch": 5.26, "learning_rate": 3.2278455691138233e-06, "loss": 0.0026, "step": 16939 }, { "epoch": 5.26, "learning_rate": 3.2274454890978196e-06, "loss": 0.0037, "step": 16940 }, { "epoch": 5.26, "learning_rate": 3.2270454090818167e-06, "loss": 0.0524, "step": 16941 }, { "epoch": 5.26, "learning_rate": 3.2266453290658133e-06, "loss": 0.004, "step": 16942 }, { "epoch": 5.26, "learning_rate": 3.2262452490498105e-06, "loss": 0.0071, "step": 16943 }, { "epoch": 5.26, "learning_rate": 3.225845169033807e-06, "loss": 0.0154, "step": 16944 }, { "epoch": 5.26, "learning_rate": 3.2254450890178043e-06, "loss": 0.0048, "step": 16945 }, { "epoch": 5.26, "learning_rate": 3.2250450090018005e-06, "loss": 0.0018, "step": 16946 }, { "epoch": 5.26, "learning_rate": 3.224644928985797e-06, "loss": 0.0078, "step": 16947 }, { "epoch": 5.26, "learning_rate": 3.2242448489697943e-06, "loss": 0.0114, "step": 16948 }, { "epoch": 5.26, "learning_rate": 3.223844768953791e-06, "loss": 0.0042, "step": 16949 }, { "epoch": 5.26, "learning_rate": 3.223444688937788e-06, "loss": 0.0175, "step": 16950 }, { "epoch": 5.26, "learning_rate": 3.223044608921785e-06, "loss": 0.0021, "step": 16951 }, { "epoch": 5.26, "learning_rate": 3.222644528905781e-06, "loss": 0.0118, "step": 16952 }, { "epoch": 5.26, "learning_rate": 3.222244448889778e-06, "loss": 0.0333, "step": 16953 }, { "epoch": 5.26, "learning_rate": 3.221844368873775e-06, "loss": 0.0057, "step": 16954 }, { "epoch": 5.26, "learning_rate": 3.221444288857772e-06, "loss": 0.0033, "step": 16955 }, { "epoch": 5.26, "learning_rate": 3.2210442088417687e-06, "loss": 0.0078, "step": 16956 }, { "epoch": 5.26, "learning_rate": 3.2206441288257658e-06, "loss": 0.0013, "step": 16957 }, { "epoch": 5.26, "learning_rate": 3.2202440488097624e-06, "loss": 0.0074, "step": 16958 }, { "epoch": 5.26, "learning_rate": 3.2198439687937587e-06, "loss": 0.0177, "step": 16959 }, { "epoch": 5.26, "learning_rate": 3.219443888777756e-06, "loss": 0.0077, "step": 16960 }, { "epoch": 5.26, "learning_rate": 3.2190438087617525e-06, "loss": 0.0066, "step": 16961 }, { "epoch": 5.26, "learning_rate": 3.2186437287457496e-06, "loss": 0.0089, "step": 16962 }, { "epoch": 5.26, "learning_rate": 3.2182436487297463e-06, "loss": 0.0035, "step": 16963 }, { "epoch": 5.27, "learning_rate": 3.2178435687137434e-06, "loss": 0.0034, "step": 16964 }, { "epoch": 5.27, "learning_rate": 3.2174434886977397e-06, "loss": 0.0062, "step": 16965 }, { "epoch": 5.27, "learning_rate": 3.2170434086817364e-06, "loss": 0.0059, "step": 16966 }, { "epoch": 5.27, "learning_rate": 3.2166433286657335e-06, "loss": 0.0023, "step": 16967 }, { "epoch": 5.27, "learning_rate": 3.21624324864973e-06, "loss": 0.0209, "step": 16968 }, { "epoch": 5.27, "learning_rate": 3.2158431686337273e-06, "loss": 0.0074, "step": 16969 }, { "epoch": 5.27, "learning_rate": 3.215443088617724e-06, "loss": 0.0233, "step": 16970 }, { "epoch": 5.27, "learning_rate": 3.2150430086017202e-06, "loss": 0.0664, "step": 16971 }, { "epoch": 5.27, "learning_rate": 3.2146429285857173e-06, "loss": 0.0014, "step": 16972 }, { "epoch": 5.27, "learning_rate": 3.214242848569714e-06, "loss": 0.0037, "step": 16973 }, { "epoch": 5.27, "learning_rate": 3.213842768553711e-06, "loss": 0.0062, "step": 16974 }, { "epoch": 5.27, "learning_rate": 3.213442688537708e-06, "loss": 0.0034, "step": 16975 }, { "epoch": 5.27, "learning_rate": 3.213042608521705e-06, "loss": 0.002, "step": 16976 }, { "epoch": 5.27, "learning_rate": 3.212642528505701e-06, "loss": 0.0029, "step": 16977 }, { "epoch": 5.27, "learning_rate": 3.212242448489698e-06, "loss": 0.0035, "step": 16978 }, { "epoch": 5.27, "learning_rate": 3.211842368473695e-06, "loss": 0.0033, "step": 16979 }, { "epoch": 5.27, "learning_rate": 3.2114422884576917e-06, "loss": 0.0137, "step": 16980 }, { "epoch": 5.27, "learning_rate": 3.2110422084416888e-06, "loss": 0.0043, "step": 16981 }, { "epoch": 5.27, "learning_rate": 3.2106421284256855e-06, "loss": 0.0074, "step": 16982 }, { "epoch": 5.27, "learning_rate": 3.2102420484096826e-06, "loss": 0.0047, "step": 16983 }, { "epoch": 5.27, "learning_rate": 3.209841968393679e-06, "loss": 0.0072, "step": 16984 }, { "epoch": 5.27, "learning_rate": 3.2094418883776755e-06, "loss": 0.0214, "step": 16985 }, { "epoch": 5.27, "learning_rate": 3.2090418083616726e-06, "loss": 0.0049, "step": 16986 }, { "epoch": 5.27, "learning_rate": 3.2086417283456693e-06, "loss": 0.0029, "step": 16987 }, { "epoch": 5.27, "learning_rate": 3.2082416483296664e-06, "loss": 0.0543, "step": 16988 }, { "epoch": 5.27, "learning_rate": 3.207841568313663e-06, "loss": 0.008, "step": 16989 }, { "epoch": 5.27, "learning_rate": 3.2074414882976594e-06, "loss": 0.0783, "step": 16990 }, { "epoch": 5.27, "learning_rate": 3.2070414082816565e-06, "loss": 0.0278, "step": 16991 }, { "epoch": 5.27, "learning_rate": 3.206641328265653e-06, "loss": 0.0091, "step": 16992 }, { "epoch": 5.27, "learning_rate": 3.2062412482496503e-06, "loss": 0.011, "step": 16993 }, { "epoch": 5.27, "learning_rate": 3.205841168233647e-06, "loss": 0.0028, "step": 16994 }, { "epoch": 5.27, "learning_rate": 3.205441088217644e-06, "loss": 0.0053, "step": 16995 }, { "epoch": 5.28, "learning_rate": 3.2050410082016403e-06, "loss": 0.005, "step": 16996 }, { "epoch": 5.28, "learning_rate": 3.204640928185637e-06, "loss": 0.0053, "step": 16997 }, { "epoch": 5.28, "learning_rate": 3.204240848169634e-06, "loss": 0.0215, "step": 16998 }, { "epoch": 5.28, "learning_rate": 3.203840768153631e-06, "loss": 0.0873, "step": 16999 }, { "epoch": 5.28, "learning_rate": 3.203440688137628e-06, "loss": 0.0048, "step": 17000 }, { "epoch": 5.28, "learning_rate": 3.2030406081216246e-06, "loss": 0.0116, "step": 17001 }, { "epoch": 5.28, "learning_rate": 3.2026405281056217e-06, "loss": 0.0032, "step": 17002 }, { "epoch": 5.28, "learning_rate": 3.202240448089618e-06, "loss": 0.003, "step": 17003 }, { "epoch": 5.28, "learning_rate": 3.2018403680736147e-06, "loss": 0.0131, "step": 17004 }, { "epoch": 5.28, "learning_rate": 3.2014402880576118e-06, "loss": 0.0052, "step": 17005 }, { "epoch": 5.28, "learning_rate": 3.2010402080416085e-06, "loss": 0.0132, "step": 17006 }, { "epoch": 5.28, "learning_rate": 3.2006401280256056e-06, "loss": 0.0019, "step": 17007 }, { "epoch": 5.28, "learning_rate": 3.2002400480096023e-06, "loss": 0.0063, "step": 17008 }, { "epoch": 5.28, "learning_rate": 3.199839967993599e-06, "loss": 0.0049, "step": 17009 }, { "epoch": 5.28, "learning_rate": 3.1994398879775956e-06, "loss": 0.0026, "step": 17010 }, { "epoch": 5.28, "learning_rate": 3.1990398079615923e-06, "loss": 0.0023, "step": 17011 }, { "epoch": 5.28, "learning_rate": 3.1986397279455894e-06, "loss": 0.0021, "step": 17012 }, { "epoch": 5.28, "learning_rate": 3.198239647929586e-06, "loss": 0.0141, "step": 17013 }, { "epoch": 5.28, "learning_rate": 3.1978395679135832e-06, "loss": 0.0017, "step": 17014 }, { "epoch": 5.28, "learning_rate": 3.1974394878975795e-06, "loss": 0.0054, "step": 17015 }, { "epoch": 5.28, "learning_rate": 3.1970394078815766e-06, "loss": 0.008, "step": 17016 }, { "epoch": 5.28, "learning_rate": 3.1966393278655733e-06, "loss": 0.0046, "step": 17017 }, { "epoch": 5.28, "learning_rate": 3.19623924784957e-06, "loss": 0.0064, "step": 17018 }, { "epoch": 5.28, "learning_rate": 3.195839167833567e-06, "loss": 0.0045, "step": 17019 }, { "epoch": 5.28, "learning_rate": 3.1954390878175638e-06, "loss": 0.0131, "step": 17020 }, { "epoch": 5.28, "learning_rate": 3.195039007801561e-06, "loss": 0.0052, "step": 17021 }, { "epoch": 5.28, "learning_rate": 3.194638927785557e-06, "loss": 0.0066, "step": 17022 }, { "epoch": 5.28, "learning_rate": 3.1942388477695543e-06, "loss": 0.0053, "step": 17023 }, { "epoch": 5.28, "learning_rate": 3.193838767753551e-06, "loss": 0.0051, "step": 17024 }, { "epoch": 5.28, "learning_rate": 3.1934386877375476e-06, "loss": 0.0036, "step": 17025 }, { "epoch": 5.28, "learning_rate": 3.1930386077215447e-06, "loss": 0.0157, "step": 17026 }, { "epoch": 5.28, "learning_rate": 3.1926385277055414e-06, "loss": 0.0068, "step": 17027 }, { "epoch": 5.29, "learning_rate": 3.192238447689538e-06, "loss": 0.006, "step": 17028 }, { "epoch": 5.29, "learning_rate": 3.191838367673535e-06, "loss": 0.0043, "step": 17029 }, { "epoch": 5.29, "learning_rate": 3.191438287657532e-06, "loss": 0.0061, "step": 17030 }, { "epoch": 5.29, "learning_rate": 3.1910382076415286e-06, "loss": 0.0074, "step": 17031 }, { "epoch": 5.29, "learning_rate": 3.1906381276255257e-06, "loss": 0.0096, "step": 17032 }, { "epoch": 5.29, "learning_rate": 3.1902380476095224e-06, "loss": 0.0052, "step": 17033 }, { "epoch": 5.29, "learning_rate": 3.1898379675935187e-06, "loss": 0.0045, "step": 17034 }, { "epoch": 5.29, "learning_rate": 3.1894378875775158e-06, "loss": 0.0344, "step": 17035 }, { "epoch": 5.29, "learning_rate": 3.1890378075615124e-06, "loss": 0.0033, "step": 17036 }, { "epoch": 5.29, "learning_rate": 3.1886377275455096e-06, "loss": 0.0034, "step": 17037 }, { "epoch": 5.29, "learning_rate": 3.1882376475295062e-06, "loss": 0.0075, "step": 17038 }, { "epoch": 5.29, "learning_rate": 3.1878375675135034e-06, "loss": 0.0044, "step": 17039 }, { "epoch": 5.29, "learning_rate": 3.1874374874975e-06, "loss": 0.0041, "step": 17040 }, { "epoch": 5.29, "learning_rate": 3.1870374074814963e-06, "loss": 0.0175, "step": 17041 }, { "epoch": 5.29, "learning_rate": 3.1866373274654934e-06, "loss": 0.0035, "step": 17042 }, { "epoch": 5.29, "learning_rate": 3.18623724744949e-06, "loss": 0.0023, "step": 17043 }, { "epoch": 5.29, "learning_rate": 3.185837167433487e-06, "loss": 0.0068, "step": 17044 }, { "epoch": 5.29, "learning_rate": 3.185437087417484e-06, "loss": 0.0208, "step": 17045 }, { "epoch": 5.29, "learning_rate": 3.185037007401481e-06, "loss": 0.0058, "step": 17046 }, { "epoch": 5.29, "learning_rate": 3.1846369273854773e-06, "loss": 0.0076, "step": 17047 }, { "epoch": 5.29, "learning_rate": 3.184236847369474e-06, "loss": 0.0055, "step": 17048 }, { "epoch": 5.29, "learning_rate": 3.183836767353471e-06, "loss": 0.0114, "step": 17049 }, { "epoch": 5.29, "learning_rate": 3.1834366873374677e-06, "loss": 0.0047, "step": 17050 }, { "epoch": 5.29, "learning_rate": 3.183036607321465e-06, "loss": 0.0521, "step": 17051 }, { "epoch": 5.29, "learning_rate": 3.1826365273054615e-06, "loss": 0.0086, "step": 17052 }, { "epoch": 5.29, "learning_rate": 3.182236447289458e-06, "loss": 0.0084, "step": 17053 }, { "epoch": 5.29, "learning_rate": 3.181836367273455e-06, "loss": 0.006, "step": 17054 }, { "epoch": 5.29, "learning_rate": 3.1814362872574516e-06, "loss": 0.0053, "step": 17055 }, { "epoch": 5.29, "learning_rate": 3.1810362072414487e-06, "loss": 0.0016, "step": 17056 }, { "epoch": 5.29, "learning_rate": 3.1806361272254454e-06, "loss": 0.0034, "step": 17057 }, { "epoch": 5.29, "learning_rate": 3.1802360472094425e-06, "loss": 0.0053, "step": 17058 }, { "epoch": 5.29, "learning_rate": 3.1798359671934388e-06, "loss": 0.0021, "step": 17059 }, { "epoch": 5.3, "learning_rate": 3.1794358871774355e-06, "loss": 0.0047, "step": 17060 }, { "epoch": 5.3, "learning_rate": 3.1790358071614326e-06, "loss": 0.0163, "step": 17061 }, { "epoch": 5.3, "learning_rate": 3.1786357271454293e-06, "loss": 0.0047, "step": 17062 }, { "epoch": 5.3, "learning_rate": 3.1782356471294264e-06, "loss": 0.0174, "step": 17063 }, { "epoch": 5.3, "learning_rate": 3.177835567113423e-06, "loss": 0.0019, "step": 17064 }, { "epoch": 5.3, "learning_rate": 3.17743548709742e-06, "loss": 0.011, "step": 17065 }, { "epoch": 5.3, "learning_rate": 3.1770354070814164e-06, "loss": 0.01, "step": 17066 }, { "epoch": 5.3, "learning_rate": 3.176635327065413e-06, "loss": 0.0047, "step": 17067 }, { "epoch": 5.3, "learning_rate": 3.1762352470494102e-06, "loss": 0.018, "step": 17068 }, { "epoch": 5.3, "learning_rate": 3.175835167033407e-06, "loss": 0.0372, "step": 17069 }, { "epoch": 5.3, "learning_rate": 3.175435087017404e-06, "loss": 0.008, "step": 17070 }, { "epoch": 5.3, "learning_rate": 3.1750350070014007e-06, "loss": 0.0042, "step": 17071 }, { "epoch": 5.3, "learning_rate": 3.174634926985397e-06, "loss": 0.0062, "step": 17072 }, { "epoch": 5.3, "learning_rate": 3.174234846969394e-06, "loss": 0.0066, "step": 17073 }, { "epoch": 5.3, "learning_rate": 3.1738347669533908e-06, "loss": 0.0062, "step": 17074 }, { "epoch": 5.3, "learning_rate": 3.173434686937388e-06, "loss": 0.0134, "step": 17075 }, { "epoch": 5.3, "learning_rate": 3.1730346069213846e-06, "loss": 0.0544, "step": 17076 }, { "epoch": 5.3, "learning_rate": 3.1726345269053817e-06, "loss": 0.0053, "step": 17077 }, { "epoch": 5.3, "learning_rate": 3.172234446889378e-06, "loss": 0.0048, "step": 17078 }, { "epoch": 5.3, "learning_rate": 3.1718343668733746e-06, "loss": 0.0052, "step": 17079 }, { "epoch": 5.3, "learning_rate": 3.1714342868573717e-06, "loss": 0.0091, "step": 17080 }, { "epoch": 5.3, "learning_rate": 3.1710342068413684e-06, "loss": 0.0032, "step": 17081 }, { "epoch": 5.3, "learning_rate": 3.1706341268253655e-06, "loss": 0.0196, "step": 17082 }, { "epoch": 5.3, "learning_rate": 3.170234046809362e-06, "loss": 0.0108, "step": 17083 }, { "epoch": 5.3, "learning_rate": 3.1698339667933593e-06, "loss": 0.0475, "step": 17084 }, { "epoch": 5.3, "learning_rate": 3.1694338867773556e-06, "loss": 0.0044, "step": 17085 }, { "epoch": 5.3, "learning_rate": 3.1690338067613523e-06, "loss": 0.0103, "step": 17086 }, { "epoch": 5.3, "learning_rate": 3.1686337267453494e-06, "loss": 0.0031, "step": 17087 }, { "epoch": 5.3, "learning_rate": 3.168233646729346e-06, "loss": 0.0101, "step": 17088 }, { "epoch": 5.3, "learning_rate": 3.167833566713343e-06, "loss": 0.0044, "step": 17089 }, { "epoch": 5.3, "learning_rate": 3.16743348669734e-06, "loss": 0.0107, "step": 17090 }, { "epoch": 5.3, "learning_rate": 3.167033406681336e-06, "loss": 0.0086, "step": 17091 }, { "epoch": 5.3, "learning_rate": 3.1666333266653332e-06, "loss": 0.0096, "step": 17092 }, { "epoch": 5.31, "learning_rate": 3.16623324664933e-06, "loss": 0.002, "step": 17093 }, { "epoch": 5.31, "learning_rate": 3.165833166633327e-06, "loss": 0.0069, "step": 17094 }, { "epoch": 5.31, "learning_rate": 3.1654330866173237e-06, "loss": 0.0034, "step": 17095 }, { "epoch": 5.31, "learning_rate": 3.165033006601321e-06, "loss": 0.0024, "step": 17096 }, { "epoch": 5.31, "learning_rate": 3.164632926585317e-06, "loss": 0.0026, "step": 17097 }, { "epoch": 5.31, "learning_rate": 3.1642328465693138e-06, "loss": 0.0163, "step": 17098 }, { "epoch": 5.31, "learning_rate": 3.163832766553311e-06, "loss": 0.0131, "step": 17099 }, { "epoch": 5.31, "learning_rate": 3.1634326865373076e-06, "loss": 0.0233, "step": 17100 }, { "epoch": 5.31, "learning_rate": 3.1630326065213047e-06, "loss": 0.0036, "step": 17101 }, { "epoch": 5.31, "learning_rate": 3.1626325265053014e-06, "loss": 0.0067, "step": 17102 }, { "epoch": 5.31, "learning_rate": 3.1622324464892985e-06, "loss": 0.0037, "step": 17103 }, { "epoch": 5.31, "learning_rate": 3.1618323664732947e-06, "loss": 0.0035, "step": 17104 }, { "epoch": 5.31, "learning_rate": 3.1614322864572914e-06, "loss": 0.0062, "step": 17105 }, { "epoch": 5.31, "learning_rate": 3.1610322064412885e-06, "loss": 0.0055, "step": 17106 }, { "epoch": 5.31, "learning_rate": 3.1606321264252852e-06, "loss": 0.0025, "step": 17107 }, { "epoch": 5.31, "learning_rate": 3.1602320464092823e-06, "loss": 0.0104, "step": 17108 }, { "epoch": 5.31, "learning_rate": 3.159831966393279e-06, "loss": 0.0042, "step": 17109 }, { "epoch": 5.31, "learning_rate": 3.1594318863772757e-06, "loss": 0.0033, "step": 17110 }, { "epoch": 5.31, "learning_rate": 3.1590318063612724e-06, "loss": 0.0102, "step": 17111 }, { "epoch": 5.31, "learning_rate": 3.158631726345269e-06, "loss": 0.0026, "step": 17112 }, { "epoch": 5.31, "learning_rate": 3.158231646329266e-06, "loss": 0.015, "step": 17113 }, { "epoch": 5.31, "learning_rate": 3.157831566313263e-06, "loss": 0.0094, "step": 17114 }, { "epoch": 5.31, "learning_rate": 3.15743148629726e-06, "loss": 0.0005, "step": 17115 }, { "epoch": 5.31, "learning_rate": 3.1570314062812562e-06, "loss": 0.009, "step": 17116 }, { "epoch": 5.31, "learning_rate": 3.1566313262652534e-06, "loss": 0.0063, "step": 17117 }, { "epoch": 5.31, "learning_rate": 3.15623124624925e-06, "loss": 0.0184, "step": 17118 }, { "epoch": 5.31, "learning_rate": 3.1558311662332467e-06, "loss": 0.0062, "step": 17119 }, { "epoch": 5.31, "learning_rate": 3.155431086217244e-06, "loss": 0.003, "step": 17120 }, { "epoch": 5.31, "learning_rate": 3.1550310062012405e-06, "loss": 0.0337, "step": 17121 }, { "epoch": 5.31, "learning_rate": 3.154630926185237e-06, "loss": 0.0036, "step": 17122 }, { "epoch": 5.31, "learning_rate": 3.154230846169234e-06, "loss": 0.0023, "step": 17123 }, { "epoch": 5.31, "learning_rate": 3.153830766153231e-06, "loss": 0.0071, "step": 17124 }, { "epoch": 5.32, "learning_rate": 3.1534306861372277e-06, "loss": 0.0063, "step": 17125 }, { "epoch": 5.32, "learning_rate": 3.1530306061212244e-06, "loss": 0.0015, "step": 17126 }, { "epoch": 5.32, "learning_rate": 3.1526305261052215e-06, "loss": 0.0077, "step": 17127 }, { "epoch": 5.32, "learning_rate": 3.152230446089218e-06, "loss": 0.0035, "step": 17128 }, { "epoch": 5.32, "learning_rate": 3.151830366073215e-06, "loss": 0.0074, "step": 17129 }, { "epoch": 5.32, "learning_rate": 3.1514302860572115e-06, "loss": 0.0071, "step": 17130 }, { "epoch": 5.32, "learning_rate": 3.1510302060412087e-06, "loss": 0.0822, "step": 17131 }, { "epoch": 5.32, "learning_rate": 3.1506301260252053e-06, "loss": 0.0071, "step": 17132 }, { "epoch": 5.32, "learning_rate": 3.150230046009202e-06, "loss": 0.0035, "step": 17133 }, { "epoch": 5.32, "learning_rate": 3.149829965993199e-06, "loss": 0.005, "step": 17134 }, { "epoch": 5.32, "learning_rate": 3.1494298859771954e-06, "loss": 0.0027, "step": 17135 }, { "epoch": 5.32, "learning_rate": 3.1490298059611925e-06, "loss": 0.0218, "step": 17136 }, { "epoch": 5.32, "learning_rate": 3.148629725945189e-06, "loss": 0.0045, "step": 17137 }, { "epoch": 5.32, "learning_rate": 3.1482296459291863e-06, "loss": 0.0072, "step": 17138 }, { "epoch": 5.32, "learning_rate": 3.147829565913183e-06, "loss": 0.0244, "step": 17139 }, { "epoch": 5.32, "learning_rate": 3.14742948589718e-06, "loss": 0.0052, "step": 17140 }, { "epoch": 5.32, "learning_rate": 3.1470294058811764e-06, "loss": 0.0045, "step": 17141 }, { "epoch": 5.32, "learning_rate": 3.146629325865173e-06, "loss": 0.0127, "step": 17142 }, { "epoch": 5.32, "learning_rate": 3.14622924584917e-06, "loss": 0.0199, "step": 17143 }, { "epoch": 5.32, "learning_rate": 3.145829165833167e-06, "loss": 0.0023, "step": 17144 }, { "epoch": 5.32, "learning_rate": 3.145429085817164e-06, "loss": 0.0196, "step": 17145 }, { "epoch": 5.32, "learning_rate": 3.1450290058011606e-06, "loss": 0.0031, "step": 17146 }, { "epoch": 5.32, "learning_rate": 3.1446289257851578e-06, "loss": 0.0014, "step": 17147 }, { "epoch": 5.32, "learning_rate": 3.144228845769154e-06, "loss": 0.0205, "step": 17148 }, { "epoch": 5.32, "learning_rate": 3.1438287657531507e-06, "loss": 0.0118, "step": 17149 }, { "epoch": 5.32, "learning_rate": 3.143428685737148e-06, "loss": 0.0096, "step": 17150 }, { "epoch": 5.32, "learning_rate": 3.1430286057211445e-06, "loss": 0.0711, "step": 17151 }, { "epoch": 5.32, "learning_rate": 3.1426285257051416e-06, "loss": 0.0055, "step": 17152 }, { "epoch": 5.32, "learning_rate": 3.1422284456891383e-06, "loss": 0.01, "step": 17153 }, { "epoch": 5.32, "learning_rate": 3.1418283656731346e-06, "loss": 0.0066, "step": 17154 }, { "epoch": 5.32, "learning_rate": 3.1414282856571317e-06, "loss": 0.0045, "step": 17155 }, { "epoch": 5.32, "learning_rate": 3.1410282056411284e-06, "loss": 0.01, "step": 17156 }, { "epoch": 5.33, "learning_rate": 3.1406281256251255e-06, "loss": 0.015, "step": 17157 }, { "epoch": 5.33, "learning_rate": 3.140228045609122e-06, "loss": 0.0484, "step": 17158 }, { "epoch": 5.33, "learning_rate": 3.1398279655931193e-06, "loss": 0.0017, "step": 17159 }, { "epoch": 5.33, "learning_rate": 3.1394278855771155e-06, "loss": 0.0133, "step": 17160 }, { "epoch": 5.33, "learning_rate": 3.139027805561112e-06, "loss": 0.0116, "step": 17161 }, { "epoch": 5.33, "learning_rate": 3.1386277255451093e-06, "loss": 0.012, "step": 17162 }, { "epoch": 5.33, "learning_rate": 3.138227645529106e-06, "loss": 0.008, "step": 17163 }, { "epoch": 5.33, "learning_rate": 3.137827565513103e-06, "loss": 0.0372, "step": 17164 }, { "epoch": 5.33, "learning_rate": 3.1374274854971e-06, "loss": 0.0053, "step": 17165 }, { "epoch": 5.33, "learning_rate": 3.137027405481097e-06, "loss": 0.01, "step": 17166 }, { "epoch": 5.33, "learning_rate": 3.136627325465093e-06, "loss": 0.003, "step": 17167 }, { "epoch": 5.33, "learning_rate": 3.13622724544909e-06, "loss": 0.0046, "step": 17168 }, { "epoch": 5.33, "learning_rate": 3.135827165433087e-06, "loss": 0.0073, "step": 17169 }, { "epoch": 5.33, "learning_rate": 3.1354270854170837e-06, "loss": 0.0157, "step": 17170 }, { "epoch": 5.33, "learning_rate": 3.1350270054010808e-06, "loss": 0.0044, "step": 17171 }, { "epoch": 5.33, "learning_rate": 3.1346269253850775e-06, "loss": 0.0058, "step": 17172 }, { "epoch": 5.33, "learning_rate": 3.1342268453690737e-06, "loss": 0.0146, "step": 17173 }, { "epoch": 5.33, "learning_rate": 3.133826765353071e-06, "loss": 0.0093, "step": 17174 }, { "epoch": 5.33, "learning_rate": 3.1334266853370675e-06, "loss": 0.0075, "step": 17175 }, { "epoch": 5.33, "learning_rate": 3.1330266053210646e-06, "loss": 0.0034, "step": 17176 }, { "epoch": 5.33, "learning_rate": 3.1326265253050613e-06, "loss": 0.0021, "step": 17177 }, { "epoch": 5.33, "learning_rate": 3.1322264452890584e-06, "loss": 0.0096, "step": 17178 }, { "epoch": 5.33, "learning_rate": 3.1318263652730547e-06, "loss": 0.003, "step": 17179 }, { "epoch": 5.33, "learning_rate": 3.1314262852570514e-06, "loss": 0.0043, "step": 17180 }, { "epoch": 5.33, "learning_rate": 3.1310262052410485e-06, "loss": 0.006, "step": 17181 }, { "epoch": 5.33, "learning_rate": 3.130626125225045e-06, "loss": 0.0072, "step": 17182 }, { "epoch": 5.33, "learning_rate": 3.1302260452090423e-06, "loss": 0.0015, "step": 17183 }, { "epoch": 5.33, "learning_rate": 3.129825965193039e-06, "loss": 0.0088, "step": 17184 }, { "epoch": 5.33, "learning_rate": 3.129425885177036e-06, "loss": 0.0048, "step": 17185 }, { "epoch": 5.33, "learning_rate": 3.1290258051610323e-06, "loss": 0.0051, "step": 17186 }, { "epoch": 5.33, "learning_rate": 3.128625725145029e-06, "loss": 0.0031, "step": 17187 }, { "epoch": 5.33, "learning_rate": 3.128225645129026e-06, "loss": 0.0226, "step": 17188 }, { "epoch": 5.34, "learning_rate": 3.127825565113023e-06, "loss": 0.0192, "step": 17189 }, { "epoch": 5.34, "learning_rate": 3.12742548509702e-06, "loss": 0.0644, "step": 17190 }, { "epoch": 5.34, "learning_rate": 3.1270254050810166e-06, "loss": 0.0045, "step": 17191 }, { "epoch": 5.34, "learning_rate": 3.126625325065013e-06, "loss": 0.0098, "step": 17192 }, { "epoch": 5.34, "learning_rate": 3.12622524504901e-06, "loss": 0.0054, "step": 17193 }, { "epoch": 5.34, "learning_rate": 3.1258251650330067e-06, "loss": 0.0015, "step": 17194 }, { "epoch": 5.34, "learning_rate": 3.1254250850170038e-06, "loss": 0.0033, "step": 17195 }, { "epoch": 5.34, "learning_rate": 3.1250250050010005e-06, "loss": 0.0077, "step": 17196 }, { "epoch": 5.34, "learning_rate": 3.1246249249849976e-06, "loss": 0.0069, "step": 17197 }, { "epoch": 5.34, "learning_rate": 3.124224844968994e-06, "loss": 0.0082, "step": 17198 }, { "epoch": 5.34, "learning_rate": 3.1238247649529905e-06, "loss": 0.004, "step": 17199 }, { "epoch": 5.34, "learning_rate": 3.1234246849369876e-06, "loss": 0.013, "step": 17200 }, { "epoch": 5.34, "learning_rate": 3.1230246049209843e-06, "loss": 0.0157, "step": 17201 }, { "epoch": 5.34, "learning_rate": 3.1226245249049814e-06, "loss": 0.007, "step": 17202 }, { "epoch": 5.34, "learning_rate": 3.122224444888978e-06, "loss": 0.0025, "step": 17203 }, { "epoch": 5.34, "learning_rate": 3.1218243648729744e-06, "loss": 0.0086, "step": 17204 }, { "epoch": 5.34, "learning_rate": 3.1214242848569715e-06, "loss": 0.005, "step": 17205 }, { "epoch": 5.34, "learning_rate": 3.121024204840968e-06, "loss": 0.0616, "step": 17206 }, { "epoch": 5.34, "learning_rate": 3.1206241248249653e-06, "loss": 0.0655, "step": 17207 }, { "epoch": 5.34, "learning_rate": 3.120224044808962e-06, "loss": 0.0109, "step": 17208 }, { "epoch": 5.34, "learning_rate": 3.119823964792959e-06, "loss": 0.0087, "step": 17209 }, { "epoch": 5.34, "learning_rate": 3.1194238847769558e-06, "loss": 0.0047, "step": 17210 }, { "epoch": 5.34, "learning_rate": 3.119023804760952e-06, "loss": 0.0063, "step": 17211 }, { "epoch": 5.34, "learning_rate": 3.118623724744949e-06, "loss": 0.0159, "step": 17212 }, { "epoch": 5.34, "learning_rate": 3.118223644728946e-06, "loss": 0.0179, "step": 17213 }, { "epoch": 5.34, "learning_rate": 3.117823564712943e-06, "loss": 0.0172, "step": 17214 }, { "epoch": 5.34, "learning_rate": 3.1174234846969396e-06, "loss": 0.0139, "step": 17215 }, { "epoch": 5.34, "learning_rate": 3.1170234046809367e-06, "loss": 0.0016, "step": 17216 }, { "epoch": 5.34, "learning_rate": 3.116623324664933e-06, "loss": 0.0184, "step": 17217 }, { "epoch": 5.34, "learning_rate": 3.11622324464893e-06, "loss": 0.0108, "step": 17218 }, { "epoch": 5.34, "learning_rate": 3.1158231646329268e-06, "loss": 0.0051, "step": 17219 }, { "epoch": 5.34, "learning_rate": 3.1154230846169235e-06, "loss": 0.0023, "step": 17220 }, { "epoch": 5.35, "learning_rate": 3.1150230046009206e-06, "loss": 0.0619, "step": 17221 }, { "epoch": 5.35, "learning_rate": 3.1146229245849173e-06, "loss": 0.0026, "step": 17222 }, { "epoch": 5.35, "learning_rate": 3.114222844568914e-06, "loss": 0.0071, "step": 17223 }, { "epoch": 5.35, "learning_rate": 3.1138227645529106e-06, "loss": 0.0084, "step": 17224 }, { "epoch": 5.35, "learning_rate": 3.1134226845369078e-06, "loss": 0.0025, "step": 17225 }, { "epoch": 5.35, "learning_rate": 3.1130226045209044e-06, "loss": 0.0051, "step": 17226 }, { "epoch": 5.35, "learning_rate": 3.112622524504901e-06, "loss": 0.0068, "step": 17227 }, { "epoch": 5.35, "learning_rate": 3.1122224444888982e-06, "loss": 0.0029, "step": 17228 }, { "epoch": 5.35, "learning_rate": 3.111822364472895e-06, "loss": 0.0035, "step": 17229 }, { "epoch": 5.35, "learning_rate": 3.1114222844568916e-06, "loss": 0.0045, "step": 17230 }, { "epoch": 5.35, "learning_rate": 3.1110222044408883e-06, "loss": 0.0079, "step": 17231 }, { "epoch": 5.35, "learning_rate": 3.1106221244248854e-06, "loss": 0.0076, "step": 17232 }, { "epoch": 5.35, "learning_rate": 3.110222044408882e-06, "loss": 0.0057, "step": 17233 }, { "epoch": 5.35, "learning_rate": 3.1098219643928788e-06, "loss": 0.0105, "step": 17234 }, { "epoch": 5.35, "learning_rate": 3.109421884376876e-06, "loss": 0.0077, "step": 17235 }, { "epoch": 5.35, "learning_rate": 3.109021804360872e-06, "loss": 0.0042, "step": 17236 }, { "epoch": 5.35, "learning_rate": 3.1086217243448693e-06, "loss": 0.006, "step": 17237 }, { "epoch": 5.35, "learning_rate": 3.108221644328866e-06, "loss": 0.0023, "step": 17238 }, { "epoch": 5.35, "learning_rate": 3.107821564312863e-06, "loss": 0.0098, "step": 17239 }, { "epoch": 5.35, "learning_rate": 3.1074214842968597e-06, "loss": 0.0034, "step": 17240 }, { "epoch": 5.35, "learning_rate": 3.1070214042808564e-06, "loss": 0.0042, "step": 17241 }, { "epoch": 5.35, "learning_rate": 3.106621324264853e-06, "loss": 0.0176, "step": 17242 }, { "epoch": 5.35, "learning_rate": 3.10622124424885e-06, "loss": 0.0039, "step": 17243 }, { "epoch": 5.35, "learning_rate": 3.105821164232847e-06, "loss": 0.0198, "step": 17244 }, { "epoch": 5.35, "learning_rate": 3.1054210842168436e-06, "loss": 0.0028, "step": 17245 }, { "epoch": 5.35, "learning_rate": 3.1050210042008407e-06, "loss": 0.011, "step": 17246 }, { "epoch": 5.35, "learning_rate": 3.1046209241848374e-06, "loss": 0.0034, "step": 17247 }, { "epoch": 5.35, "learning_rate": 3.1042208441688345e-06, "loss": 0.0421, "step": 17248 }, { "epoch": 5.35, "learning_rate": 3.1038207641528308e-06, "loss": 0.0058, "step": 17249 }, { "epoch": 5.35, "learning_rate": 3.1034206841368275e-06, "loss": 0.0186, "step": 17250 }, { "epoch": 5.35, "learning_rate": 3.1030206041208246e-06, "loss": 0.0054, "step": 17251 }, { "epoch": 5.35, "learning_rate": 3.1026205241048212e-06, "loss": 0.0099, "step": 17252 }, { "epoch": 5.35, "learning_rate": 3.1022204440888184e-06, "loss": 0.0046, "step": 17253 }, { "epoch": 5.36, "learning_rate": 3.101820364072815e-06, "loss": 0.008, "step": 17254 }, { "epoch": 5.36, "learning_rate": 3.1014202840568113e-06, "loss": 0.0057, "step": 17255 }, { "epoch": 5.36, "learning_rate": 3.1010202040408084e-06, "loss": 0.006, "step": 17256 }, { "epoch": 5.36, "learning_rate": 3.100620124024805e-06, "loss": 0.0084, "step": 17257 }, { "epoch": 5.36, "learning_rate": 3.100220044008802e-06, "loss": 0.0021, "step": 17258 }, { "epoch": 5.36, "learning_rate": 3.099819963992799e-06, "loss": 0.0186, "step": 17259 }, { "epoch": 5.36, "learning_rate": 3.099419883976796e-06, "loss": 0.0396, "step": 17260 }, { "epoch": 5.36, "learning_rate": 3.0990198039607923e-06, "loss": 0.0013, "step": 17261 }, { "epoch": 5.36, "learning_rate": 3.098619723944789e-06, "loss": 0.0033, "step": 17262 }, { "epoch": 5.36, "learning_rate": 3.098219643928786e-06, "loss": 0.015, "step": 17263 }, { "epoch": 5.36, "learning_rate": 3.0978195639127828e-06, "loss": 0.0027, "step": 17264 }, { "epoch": 5.36, "learning_rate": 3.09741948389678e-06, "loss": 0.0282, "step": 17265 }, { "epoch": 5.36, "learning_rate": 3.0970194038807765e-06, "loss": 0.0041, "step": 17266 }, { "epoch": 5.36, "learning_rate": 3.096619323864773e-06, "loss": 0.0523, "step": 17267 }, { "epoch": 5.36, "learning_rate": 3.09621924384877e-06, "loss": 0.0172, "step": 17268 }, { "epoch": 5.36, "learning_rate": 3.0958191638327666e-06, "loss": 0.0025, "step": 17269 }, { "epoch": 5.36, "learning_rate": 3.0954190838167637e-06, "loss": 0.0042, "step": 17270 }, { "epoch": 5.36, "learning_rate": 3.0950190038007604e-06, "loss": 0.0218, "step": 17271 }, { "epoch": 5.36, "learning_rate": 3.0946189237847575e-06, "loss": 0.0022, "step": 17272 }, { "epoch": 5.36, "learning_rate": 3.094218843768754e-06, "loss": 0.003, "step": 17273 }, { "epoch": 5.36, "learning_rate": 3.0938187637527505e-06, "loss": 0.0022, "step": 17274 }, { "epoch": 5.36, "learning_rate": 3.0934186837367476e-06, "loss": 0.0038, "step": 17275 }, { "epoch": 5.36, "learning_rate": 3.0930186037207443e-06, "loss": 0.01, "step": 17276 }, { "epoch": 5.36, "learning_rate": 3.0926185237047414e-06, "loss": 0.0562, "step": 17277 }, { "epoch": 5.36, "learning_rate": 3.092218443688738e-06, "loss": 0.0154, "step": 17278 }, { "epoch": 5.36, "learning_rate": 3.091818363672735e-06, "loss": 0.0182, "step": 17279 }, { "epoch": 5.36, "learning_rate": 3.0914182836567314e-06, "loss": 0.0197, "step": 17280 }, { "epoch": 5.36, "learning_rate": 3.091018203640728e-06, "loss": 0.0065, "step": 17281 }, { "epoch": 5.36, "learning_rate": 3.0906181236247252e-06, "loss": 0.0019, "step": 17282 }, { "epoch": 5.36, "learning_rate": 3.090218043608722e-06, "loss": 0.0021, "step": 17283 }, { "epoch": 5.36, "learning_rate": 3.089817963592719e-06, "loss": 0.0053, "step": 17284 }, { "epoch": 5.36, "learning_rate": 3.0894178835767157e-06, "loss": 0.0383, "step": 17285 }, { "epoch": 5.37, "learning_rate": 3.089017803560712e-06, "loss": 0.0063, "step": 17286 }, { "epoch": 5.37, "learning_rate": 3.088617723544709e-06, "loss": 0.006, "step": 17287 }, { "epoch": 5.37, "learning_rate": 3.0882176435287058e-06, "loss": 0.0125, "step": 17288 }, { "epoch": 5.37, "learning_rate": 3.087817563512703e-06, "loss": 0.0178, "step": 17289 }, { "epoch": 5.37, "learning_rate": 3.0874174834966996e-06, "loss": 0.009, "step": 17290 }, { "epoch": 5.37, "learning_rate": 3.0870174034806967e-06, "loss": 0.0035, "step": 17291 }, { "epoch": 5.37, "learning_rate": 3.0866173234646934e-06, "loss": 0.0123, "step": 17292 }, { "epoch": 5.37, "learning_rate": 3.0862172434486896e-06, "loss": 0.0021, "step": 17293 }, { "epoch": 5.37, "learning_rate": 3.0858171634326867e-06, "loss": 0.0079, "step": 17294 }, { "epoch": 5.37, "learning_rate": 3.0854170834166834e-06, "loss": 0.0087, "step": 17295 }, { "epoch": 5.37, "learning_rate": 3.0850170034006805e-06, "loss": 0.0022, "step": 17296 }, { "epoch": 5.37, "learning_rate": 3.084616923384677e-06, "loss": 0.0088, "step": 17297 }, { "epoch": 5.37, "learning_rate": 3.0842168433686743e-06, "loss": 0.0132, "step": 17298 }, { "epoch": 5.37, "learning_rate": 3.0838167633526706e-06, "loss": 0.0049, "step": 17299 }, { "epoch": 5.37, "learning_rate": 3.0834166833366673e-06, "loss": 0.0031, "step": 17300 }, { "epoch": 5.37, "learning_rate": 3.0830166033206644e-06, "loss": 0.0625, "step": 17301 }, { "epoch": 5.37, "learning_rate": 3.082616523304661e-06, "loss": 0.0084, "step": 17302 }, { "epoch": 5.37, "learning_rate": 3.082216443288658e-06, "loss": 0.022, "step": 17303 }, { "epoch": 5.37, "learning_rate": 3.081816363272655e-06, "loss": 0.0147, "step": 17304 }, { "epoch": 5.37, "learning_rate": 3.081416283256651e-06, "loss": 0.0032, "step": 17305 }, { "epoch": 5.37, "learning_rate": 3.0810162032406482e-06, "loss": 0.0098, "step": 17306 }, { "epoch": 5.37, "learning_rate": 3.080616123224645e-06, "loss": 0.0039, "step": 17307 }, { "epoch": 5.37, "learning_rate": 3.080216043208642e-06, "loss": 0.0045, "step": 17308 }, { "epoch": 5.37, "learning_rate": 3.0798159631926387e-06, "loss": 0.0118, "step": 17309 }, { "epoch": 5.37, "learning_rate": 3.079415883176636e-06, "loss": 0.0026, "step": 17310 }, { "epoch": 5.37, "learning_rate": 3.0790158031606325e-06, "loss": 0.0018, "step": 17311 }, { "epoch": 5.37, "learning_rate": 3.0786157231446288e-06, "loss": 0.0052, "step": 17312 }, { "epoch": 5.37, "learning_rate": 3.078215643128626e-06, "loss": 0.0034, "step": 17313 }, { "epoch": 5.37, "learning_rate": 3.0778155631126226e-06, "loss": 0.0083, "step": 17314 }, { "epoch": 5.37, "learning_rate": 3.0774154830966197e-06, "loss": 0.0081, "step": 17315 }, { "epoch": 5.37, "learning_rate": 3.0770154030806164e-06, "loss": 0.0134, "step": 17316 }, { "epoch": 5.37, "learning_rate": 3.0766153230646135e-06, "loss": 0.014, "step": 17317 }, { "epoch": 5.38, "learning_rate": 3.0762152430486097e-06, "loss": 0.0087, "step": 17318 }, { "epoch": 5.38, "learning_rate": 3.0758151630326064e-06, "loss": 0.0031, "step": 17319 }, { "epoch": 5.38, "learning_rate": 3.0754150830166035e-06, "loss": 0.0072, "step": 17320 }, { "epoch": 5.38, "learning_rate": 3.0750150030006002e-06, "loss": 0.0068, "step": 17321 }, { "epoch": 5.38, "learning_rate": 3.0746149229845973e-06, "loss": 0.0053, "step": 17322 }, { "epoch": 5.38, "learning_rate": 3.074214842968594e-06, "loss": 0.0017, "step": 17323 }, { "epoch": 5.38, "learning_rate": 3.0738147629525907e-06, "loss": 0.0063, "step": 17324 }, { "epoch": 5.38, "learning_rate": 3.0734146829365874e-06, "loss": 0.0125, "step": 17325 }, { "epoch": 5.38, "learning_rate": 3.073014602920584e-06, "loss": 0.0126, "step": 17326 }, { "epoch": 5.38, "learning_rate": 3.072614522904581e-06, "loss": 0.0077, "step": 17327 }, { "epoch": 5.38, "learning_rate": 3.072214442888578e-06, "loss": 0.0216, "step": 17328 }, { "epoch": 5.38, "learning_rate": 3.071814362872575e-06, "loss": 0.0136, "step": 17329 }, { "epoch": 5.38, "learning_rate": 3.0714142828565717e-06, "loss": 0.0062, "step": 17330 }, { "epoch": 5.38, "learning_rate": 3.0710142028405684e-06, "loss": 0.0042, "step": 17331 }, { "epoch": 5.38, "learning_rate": 3.070614122824565e-06, "loss": 0.0046, "step": 17332 }, { "epoch": 5.38, "learning_rate": 3.070214042808562e-06, "loss": 0.0094, "step": 17333 }, { "epoch": 5.38, "learning_rate": 3.069813962792559e-06, "loss": 0.0031, "step": 17334 }, { "epoch": 5.38, "learning_rate": 3.0694138827765555e-06, "loss": 0.0084, "step": 17335 }, { "epoch": 5.38, "learning_rate": 3.0690138027605526e-06, "loss": 0.0052, "step": 17336 }, { "epoch": 5.38, "learning_rate": 3.068613722744549e-06, "loss": 0.0037, "step": 17337 }, { "epoch": 5.38, "learning_rate": 3.068213642728546e-06, "loss": 0.0082, "step": 17338 }, { "epoch": 5.38, "learning_rate": 3.0678135627125427e-06, "loss": 0.1418, "step": 17339 }, { "epoch": 5.38, "learning_rate": 3.06741348269654e-06, "loss": 0.006, "step": 17340 }, { "epoch": 5.38, "learning_rate": 3.0670134026805365e-06, "loss": 0.0108, "step": 17341 }, { "epoch": 5.38, "learning_rate": 3.066613322664533e-06, "loss": 0.0046, "step": 17342 }, { "epoch": 5.38, "learning_rate": 3.06621324264853e-06, "loss": 0.0167, "step": 17343 }, { "epoch": 5.38, "learning_rate": 3.0658131626325265e-06, "loss": 0.0038, "step": 17344 }, { "epoch": 5.38, "learning_rate": 3.0654130826165237e-06, "loss": 0.0022, "step": 17345 }, { "epoch": 5.38, "learning_rate": 3.0650130026005203e-06, "loss": 0.007, "step": 17346 }, { "epoch": 5.38, "learning_rate": 3.0646129225845175e-06, "loss": 0.017, "step": 17347 }, { "epoch": 5.38, "learning_rate": 3.064212842568514e-06, "loss": 0.0041, "step": 17348 }, { "epoch": 5.38, "learning_rate": 3.0638127625525104e-06, "loss": 0.0097, "step": 17349 }, { "epoch": 5.39, "learning_rate": 3.0634126825365075e-06, "loss": 0.0032, "step": 17350 }, { "epoch": 5.39, "learning_rate": 3.063012602520504e-06, "loss": 0.0054, "step": 17351 }, { "epoch": 5.39, "learning_rate": 3.0626125225045013e-06, "loss": 0.0024, "step": 17352 }, { "epoch": 5.39, "learning_rate": 3.062212442488498e-06, "loss": 0.0045, "step": 17353 }, { "epoch": 5.39, "learning_rate": 3.061812362472495e-06, "loss": 0.0494, "step": 17354 }, { "epoch": 5.39, "learning_rate": 3.061412282456492e-06, "loss": 0.0091, "step": 17355 }, { "epoch": 5.39, "learning_rate": 3.061012202440488e-06, "loss": 0.0041, "step": 17356 }, { "epoch": 5.39, "learning_rate": 3.060612122424485e-06, "loss": 0.0471, "step": 17357 }, { "epoch": 5.39, "learning_rate": 3.060212042408482e-06, "loss": 0.0128, "step": 17358 }, { "epoch": 5.39, "learning_rate": 3.059811962392479e-06, "loss": 0.0095, "step": 17359 }, { "epoch": 5.39, "learning_rate": 3.0594118823764756e-06, "loss": 0.0096, "step": 17360 }, { "epoch": 5.39, "learning_rate": 3.0590118023604728e-06, "loss": 0.0099, "step": 17361 }, { "epoch": 5.39, "learning_rate": 3.058611722344469e-06, "loss": 0.0027, "step": 17362 }, { "epoch": 5.39, "learning_rate": 3.0582116423284657e-06, "loss": 0.0044, "step": 17363 }, { "epoch": 5.39, "learning_rate": 3.057811562312463e-06, "loss": 0.0041, "step": 17364 }, { "epoch": 5.39, "learning_rate": 3.0574114822964595e-06, "loss": 0.0199, "step": 17365 }, { "epoch": 5.39, "learning_rate": 3.0570114022804566e-06, "loss": 0.0144, "step": 17366 }, { "epoch": 5.39, "learning_rate": 3.0566113222644533e-06, "loss": 0.0074, "step": 17367 }, { "epoch": 5.39, "learning_rate": 3.0562112422484496e-06, "loss": 0.0025, "step": 17368 }, { "epoch": 5.39, "learning_rate": 3.0558111622324467e-06, "loss": 0.1126, "step": 17369 }, { "epoch": 5.39, "learning_rate": 3.0554110822164434e-06, "loss": 0.0059, "step": 17370 }, { "epoch": 5.39, "learning_rate": 3.0550110022004405e-06, "loss": 0.0044, "step": 17371 }, { "epoch": 5.39, "learning_rate": 3.054610922184437e-06, "loss": 0.0035, "step": 17372 }, { "epoch": 5.39, "learning_rate": 3.0542108421684343e-06, "loss": 0.0115, "step": 17373 }, { "epoch": 5.39, "learning_rate": 3.053810762152431e-06, "loss": 0.0086, "step": 17374 }, { "epoch": 5.39, "learning_rate": 3.053410682136427e-06, "loss": 0.0326, "step": 17375 }, { "epoch": 5.39, "learning_rate": 3.0530106021204243e-06, "loss": 0.0148, "step": 17376 }, { "epoch": 5.39, "learning_rate": 3.052610522104421e-06, "loss": 0.0032, "step": 17377 }, { "epoch": 5.39, "learning_rate": 3.052210442088418e-06, "loss": 0.0095, "step": 17378 }, { "epoch": 5.39, "learning_rate": 3.051810362072415e-06, "loss": 0.0022, "step": 17379 }, { "epoch": 5.39, "learning_rate": 3.051410282056412e-06, "loss": 0.0124, "step": 17380 }, { "epoch": 5.39, "learning_rate": 3.051010202040408e-06, "loss": 0.0021, "step": 17381 }, { "epoch": 5.39, "learning_rate": 3.050610122024405e-06, "loss": 0.0028, "step": 17382 }, { "epoch": 5.4, "learning_rate": 3.050210042008402e-06, "loss": 0.0055, "step": 17383 }, { "epoch": 5.4, "learning_rate": 3.0498099619923987e-06, "loss": 0.0068, "step": 17384 }, { "epoch": 5.4, "learning_rate": 3.0494098819763958e-06, "loss": 0.0161, "step": 17385 }, { "epoch": 5.4, "learning_rate": 3.0490098019603925e-06, "loss": 0.0238, "step": 17386 }, { "epoch": 5.4, "learning_rate": 3.0486097219443887e-06, "loss": 0.0039, "step": 17387 }, { "epoch": 5.4, "learning_rate": 3.048209641928386e-06, "loss": 0.006, "step": 17388 }, { "epoch": 5.4, "learning_rate": 3.0478095619123825e-06, "loss": 0.0079, "step": 17389 }, { "epoch": 5.4, "learning_rate": 3.0474094818963796e-06, "loss": 0.0062, "step": 17390 }, { "epoch": 5.4, "learning_rate": 3.0470094018803763e-06, "loss": 0.0283, "step": 17391 }, { "epoch": 5.4, "learning_rate": 3.0466093218643734e-06, "loss": 0.0038, "step": 17392 }, { "epoch": 5.4, "learning_rate": 3.04620924184837e-06, "loss": 0.0109, "step": 17393 }, { "epoch": 5.4, "learning_rate": 3.0458091618323664e-06, "loss": 0.0036, "step": 17394 }, { "epoch": 5.4, "learning_rate": 3.0454090818163635e-06, "loss": 0.0071, "step": 17395 }, { "epoch": 5.4, "learning_rate": 3.04500900180036e-06, "loss": 0.0342, "step": 17396 }, { "epoch": 5.4, "learning_rate": 3.0446089217843573e-06, "loss": 0.0184, "step": 17397 }, { "epoch": 5.4, "learning_rate": 3.044208841768354e-06, "loss": 0.0031, "step": 17398 }, { "epoch": 5.4, "learning_rate": 3.043808761752351e-06, "loss": 0.0035, "step": 17399 }, { "epoch": 5.4, "learning_rate": 3.0434086817363473e-06, "loss": 0.0042, "step": 17400 }, { "epoch": 5.4, "learning_rate": 3.043008601720344e-06, "loss": 0.0016, "step": 17401 }, { "epoch": 5.4, "learning_rate": 3.042608521704341e-06, "loss": 0.009, "step": 17402 }, { "epoch": 5.4, "learning_rate": 3.042208441688338e-06, "loss": 0.002, "step": 17403 }, { "epoch": 5.4, "learning_rate": 3.041808361672335e-06, "loss": 0.0076, "step": 17404 }, { "epoch": 5.4, "learning_rate": 3.0414082816563316e-06, "loss": 0.0096, "step": 17405 }, { "epoch": 5.4, "learning_rate": 3.041008201640328e-06, "loss": 0.0148, "step": 17406 }, { "epoch": 5.4, "learning_rate": 3.040608121624325e-06, "loss": 0.0027, "step": 17407 }, { "epoch": 5.4, "learning_rate": 3.0402080416083217e-06, "loss": 0.0123, "step": 17408 }, { "epoch": 5.4, "learning_rate": 3.0398079615923188e-06, "loss": 0.0172, "step": 17409 }, { "epoch": 5.4, "learning_rate": 3.0394078815763155e-06, "loss": 0.0028, "step": 17410 }, { "epoch": 5.4, "learning_rate": 3.0390078015603126e-06, "loss": 0.0055, "step": 17411 }, { "epoch": 5.4, "learning_rate": 3.0386077215443093e-06, "loss": 0.0167, "step": 17412 }, { "epoch": 5.4, "learning_rate": 3.0382076415283055e-06, "loss": 0.0048, "step": 17413 }, { "epoch": 5.4, "learning_rate": 3.0378075615123026e-06, "loss": 0.0019, "step": 17414 }, { "epoch": 5.41, "learning_rate": 3.0374074814962993e-06, "loss": 0.0075, "step": 17415 }, { "epoch": 5.41, "learning_rate": 3.0370074014802964e-06, "loss": 0.0089, "step": 17416 }, { "epoch": 5.41, "learning_rate": 3.036607321464293e-06, "loss": 0.0075, "step": 17417 }, { "epoch": 5.41, "learning_rate": 3.0362072414482902e-06, "loss": 0.0067, "step": 17418 }, { "epoch": 5.41, "learning_rate": 3.0358071614322865e-06, "loss": 0.0707, "step": 17419 }, { "epoch": 5.41, "learning_rate": 3.035407081416283e-06, "loss": 0.0034, "step": 17420 }, { "epoch": 5.41, "learning_rate": 3.0350070014002803e-06, "loss": 0.0022, "step": 17421 }, { "epoch": 5.41, "learning_rate": 3.034606921384277e-06, "loss": 0.0064, "step": 17422 }, { "epoch": 5.41, "learning_rate": 3.034206841368274e-06, "loss": 0.0032, "step": 17423 }, { "epoch": 5.41, "learning_rate": 3.0338067613522708e-06, "loss": 0.0045, "step": 17424 }, { "epoch": 5.41, "learning_rate": 3.0334066813362675e-06, "loss": 0.0117, "step": 17425 }, { "epoch": 5.41, "learning_rate": 3.033006601320264e-06, "loss": 0.0025, "step": 17426 }, { "epoch": 5.41, "learning_rate": 3.032606521304261e-06, "loss": 0.013, "step": 17427 }, { "epoch": 5.41, "learning_rate": 3.032206441288258e-06, "loss": 0.011, "step": 17428 }, { "epoch": 5.41, "learning_rate": 3.0318063612722546e-06, "loss": 0.0033, "step": 17429 }, { "epoch": 5.41, "learning_rate": 3.0314062812562517e-06, "loss": 0.0294, "step": 17430 }, { "epoch": 5.41, "learning_rate": 3.031006201240248e-06, "loss": 0.0081, "step": 17431 }, { "epoch": 5.41, "learning_rate": 3.030606121224245e-06, "loss": 0.0175, "step": 17432 }, { "epoch": 5.41, "learning_rate": 3.030206041208242e-06, "loss": 0.0102, "step": 17433 }, { "epoch": 5.41, "learning_rate": 3.0298059611922385e-06, "loss": 0.0028, "step": 17434 }, { "epoch": 5.41, "learning_rate": 3.0294058811762356e-06, "loss": 0.0584, "step": 17435 }, { "epoch": 5.41, "learning_rate": 3.0290058011602323e-06, "loss": 0.0051, "step": 17436 }, { "epoch": 5.41, "learning_rate": 3.0286057211442294e-06, "loss": 0.0451, "step": 17437 }, { "epoch": 5.41, "learning_rate": 3.0282056411282256e-06, "loss": 0.0158, "step": 17438 }, { "epoch": 5.41, "learning_rate": 3.0278055611122228e-06, "loss": 0.008, "step": 17439 }, { "epoch": 5.41, "learning_rate": 3.0274054810962194e-06, "loss": 0.0302, "step": 17440 }, { "epoch": 5.41, "learning_rate": 3.0270054010802166e-06, "loss": 0.0099, "step": 17441 }, { "epoch": 5.41, "learning_rate": 3.0266053210642132e-06, "loss": 0.0069, "step": 17442 }, { "epoch": 5.41, "learning_rate": 3.02620524104821e-06, "loss": 0.0019, "step": 17443 }, { "epoch": 5.41, "learning_rate": 3.0258051610322066e-06, "loss": 0.0018, "step": 17444 }, { "epoch": 5.41, "learning_rate": 3.0254050810162033e-06, "loss": 0.0455, "step": 17445 }, { "epoch": 5.41, "learning_rate": 3.0250050010002004e-06, "loss": 0.0116, "step": 17446 }, { "epoch": 5.42, "learning_rate": 3.024604920984197e-06, "loss": 0.0026, "step": 17447 }, { "epoch": 5.42, "learning_rate": 3.024204840968194e-06, "loss": 0.0053, "step": 17448 }, { "epoch": 5.42, "learning_rate": 3.023804760952191e-06, "loss": 0.0109, "step": 17449 }, { "epoch": 5.42, "learning_rate": 3.023404680936187e-06, "loss": 0.002, "step": 17450 }, { "epoch": 5.42, "learning_rate": 3.0230046009201843e-06, "loss": 0.002, "step": 17451 }, { "epoch": 5.42, "learning_rate": 3.022604520904181e-06, "loss": 0.0043, "step": 17452 }, { "epoch": 5.42, "learning_rate": 3.022204440888178e-06, "loss": 0.0068, "step": 17453 }, { "epoch": 5.42, "learning_rate": 3.0218043608721747e-06, "loss": 0.0026, "step": 17454 }, { "epoch": 5.42, "learning_rate": 3.021404280856172e-06, "loss": 0.0191, "step": 17455 }, { "epoch": 5.42, "learning_rate": 3.0210042008401685e-06, "loss": 0.0101, "step": 17456 }, { "epoch": 5.42, "learning_rate": 3.020604120824165e-06, "loss": 0.0418, "step": 17457 }, { "epoch": 5.42, "learning_rate": 3.020204040808162e-06, "loss": 0.0047, "step": 17458 }, { "epoch": 5.42, "learning_rate": 3.0198039607921586e-06, "loss": 0.0097, "step": 17459 }, { "epoch": 5.42, "learning_rate": 3.0194038807761557e-06, "loss": 0.005, "step": 17460 }, { "epoch": 5.42, "learning_rate": 3.0190038007601524e-06, "loss": 0.0082, "step": 17461 }, { "epoch": 5.42, "learning_rate": 3.0186037207441495e-06, "loss": 0.0191, "step": 17462 }, { "epoch": 5.42, "learning_rate": 3.0182036407281458e-06, "loss": 0.0076, "step": 17463 }, { "epoch": 5.42, "learning_rate": 3.0178035607121425e-06, "loss": 0.0051, "step": 17464 }, { "epoch": 5.42, "learning_rate": 3.0174034806961396e-06, "loss": 0.0123, "step": 17465 }, { "epoch": 5.42, "learning_rate": 3.0170034006801363e-06, "loss": 0.0159, "step": 17466 }, { "epoch": 5.42, "learning_rate": 3.0166033206641334e-06, "loss": 0.0035, "step": 17467 }, { "epoch": 5.42, "learning_rate": 3.01620324064813e-06, "loss": 0.0055, "step": 17468 }, { "epoch": 5.42, "learning_rate": 3.0158031606321263e-06, "loss": 0.0078, "step": 17469 }, { "epoch": 5.42, "learning_rate": 3.0154030806161234e-06, "loss": 0.0063, "step": 17470 }, { "epoch": 5.42, "learning_rate": 3.01500300060012e-06, "loss": 0.0084, "step": 17471 }, { "epoch": 5.42, "learning_rate": 3.0146029205841172e-06, "loss": 0.0129, "step": 17472 }, { "epoch": 5.42, "learning_rate": 3.014202840568114e-06, "loss": 0.0048, "step": 17473 }, { "epoch": 5.42, "learning_rate": 3.013802760552111e-06, "loss": 0.003, "step": 17474 }, { "epoch": 5.42, "learning_rate": 3.0134026805361077e-06, "loss": 0.004, "step": 17475 }, { "epoch": 5.42, "learning_rate": 3.013002600520104e-06, "loss": 0.0047, "step": 17476 }, { "epoch": 5.42, "learning_rate": 3.012602520504101e-06, "loss": 0.0076, "step": 17477 }, { "epoch": 5.42, "learning_rate": 3.0122024404880978e-06, "loss": 0.0077, "step": 17478 }, { "epoch": 5.43, "learning_rate": 3.011802360472095e-06, "loss": 0.0242, "step": 17479 }, { "epoch": 5.43, "learning_rate": 3.0114022804560916e-06, "loss": 0.0055, "step": 17480 }, { "epoch": 5.43, "learning_rate": 3.0110022004400887e-06, "loss": 0.0039, "step": 17481 }, { "epoch": 5.43, "learning_rate": 3.010602120424085e-06, "loss": 0.0075, "step": 17482 }, { "epoch": 5.43, "learning_rate": 3.0102020404080816e-06, "loss": 0.0109, "step": 17483 }, { "epoch": 5.43, "learning_rate": 3.0098019603920787e-06, "loss": 0.0019, "step": 17484 }, { "epoch": 5.43, "learning_rate": 3.0094018803760754e-06, "loss": 0.0069, "step": 17485 }, { "epoch": 5.43, "learning_rate": 3.0090018003600725e-06, "loss": 0.0035, "step": 17486 }, { "epoch": 5.43, "learning_rate": 3.008601720344069e-06, "loss": 0.0063, "step": 17487 }, { "epoch": 5.43, "learning_rate": 3.0082016403280655e-06, "loss": 0.0019, "step": 17488 }, { "epoch": 5.43, "learning_rate": 3.0078015603120626e-06, "loss": 0.0635, "step": 17489 }, { "epoch": 5.43, "learning_rate": 3.0074014802960593e-06, "loss": 0.0143, "step": 17490 }, { "epoch": 5.43, "learning_rate": 3.0070014002800564e-06, "loss": 0.0081, "step": 17491 }, { "epoch": 5.43, "learning_rate": 3.006601320264053e-06, "loss": 0.0045, "step": 17492 }, { "epoch": 5.43, "learning_rate": 3.00620124024805e-06, "loss": 0.0067, "step": 17493 }, { "epoch": 5.43, "learning_rate": 3.0058011602320464e-06, "loss": 0.0038, "step": 17494 }, { "epoch": 5.43, "learning_rate": 3.005401080216043e-06, "loss": 0.0022, "step": 17495 }, { "epoch": 5.43, "learning_rate": 3.0050010002000402e-06, "loss": 0.008, "step": 17496 }, { "epoch": 5.43, "learning_rate": 3.004600920184037e-06, "loss": 0.0023, "step": 17497 }, { "epoch": 5.43, "learning_rate": 3.004200840168034e-06, "loss": 0.0062, "step": 17498 }, { "epoch": 5.43, "learning_rate": 3.0038007601520307e-06, "loss": 0.0021, "step": 17499 }, { "epoch": 5.43, "learning_rate": 3.003400680136028e-06, "loss": 0.0016, "step": 17500 }, { "epoch": 5.43, "learning_rate": 3.003000600120024e-06, "loss": 0.0061, "step": 17501 }, { "epoch": 5.43, "learning_rate": 3.0026005201040208e-06, "loss": 0.0108, "step": 17502 }, { "epoch": 5.43, "learning_rate": 3.002200440088018e-06, "loss": 0.0021, "step": 17503 }, { "epoch": 5.43, "learning_rate": 3.0018003600720146e-06, "loss": 0.0792, "step": 17504 }, { "epoch": 5.43, "learning_rate": 3.0014002800560117e-06, "loss": 0.0107, "step": 17505 }, { "epoch": 5.43, "learning_rate": 3.0010002000400084e-06, "loss": 0.0235, "step": 17506 }, { "epoch": 5.43, "learning_rate": 3.0006001200240046e-06, "loss": 0.0128, "step": 17507 }, { "epoch": 5.43, "learning_rate": 3.0002000400080017e-06, "loss": 0.0071, "step": 17508 }, { "epoch": 5.43, "learning_rate": 2.9997999599919984e-06, "loss": 0.0286, "step": 17509 }, { "epoch": 5.43, "learning_rate": 2.9993998799759955e-06, "loss": 0.0056, "step": 17510 }, { "epoch": 5.44, "learning_rate": 2.9989997999599922e-06, "loss": 0.0011, "step": 17511 }, { "epoch": 5.44, "learning_rate": 2.9985997199439893e-06, "loss": 0.0064, "step": 17512 }, { "epoch": 5.44, "learning_rate": 2.9981996399279856e-06, "loss": 0.0019, "step": 17513 }, { "epoch": 5.44, "learning_rate": 2.9977995599119823e-06, "loss": 0.0036, "step": 17514 }, { "epoch": 5.44, "learning_rate": 2.9973994798959794e-06, "loss": 0.0125, "step": 17515 }, { "epoch": 5.44, "learning_rate": 2.996999399879976e-06, "loss": 0.0193, "step": 17516 }, { "epoch": 5.44, "learning_rate": 2.996599319863973e-06, "loss": 0.0036, "step": 17517 }, { "epoch": 5.44, "learning_rate": 2.99619923984797e-06, "loss": 0.0056, "step": 17518 }, { "epoch": 5.44, "learning_rate": 2.995799159831967e-06, "loss": 0.0043, "step": 17519 }, { "epoch": 5.44, "learning_rate": 2.9953990798159632e-06, "loss": 0.0074, "step": 17520 }, { "epoch": 5.44, "learning_rate": 2.99499899979996e-06, "loss": 0.0508, "step": 17521 }, { "epoch": 5.44, "learning_rate": 2.994598919783957e-06, "loss": 0.0107, "step": 17522 }, { "epoch": 5.44, "learning_rate": 2.9941988397679537e-06, "loss": 0.0213, "step": 17523 }, { "epoch": 5.44, "learning_rate": 2.993798759751951e-06, "loss": 0.0023, "step": 17524 }, { "epoch": 5.44, "learning_rate": 2.9933986797359475e-06, "loss": 0.0017, "step": 17525 }, { "epoch": 5.44, "learning_rate": 2.992998599719944e-06, "loss": 0.016, "step": 17526 }, { "epoch": 5.44, "learning_rate": 2.992598519703941e-06, "loss": 0.0042, "step": 17527 }, { "epoch": 5.44, "learning_rate": 2.9921984396879376e-06, "loss": 0.0223, "step": 17528 }, { "epoch": 5.44, "learning_rate": 2.9917983596719347e-06, "loss": 0.0022, "step": 17529 }, { "epoch": 5.44, "learning_rate": 2.9913982796559314e-06, "loss": 0.0033, "step": 17530 }, { "epoch": 5.44, "learning_rate": 2.9909981996399285e-06, "loss": 0.0039, "step": 17531 }, { "epoch": 5.44, "learning_rate": 2.9905981196239247e-06, "loss": 0.0128, "step": 17532 }, { "epoch": 5.44, "learning_rate": 2.990198039607922e-06, "loss": 0.005, "step": 17533 }, { "epoch": 5.44, "learning_rate": 2.9897979595919185e-06, "loss": 0.0145, "step": 17534 }, { "epoch": 5.44, "learning_rate": 2.9893978795759152e-06, "loss": 0.0527, "step": 17535 }, { "epoch": 5.44, "learning_rate": 2.9889977995599123e-06, "loss": 0.0189, "step": 17536 }, { "epoch": 5.44, "learning_rate": 2.988597719543909e-06, "loss": 0.0104, "step": 17537 }, { "epoch": 5.44, "learning_rate": 2.988197639527906e-06, "loss": 0.0482, "step": 17538 }, { "epoch": 5.44, "learning_rate": 2.9877975595119024e-06, "loss": 0.0108, "step": 17539 }, { "epoch": 5.44, "learning_rate": 2.9873974794958995e-06, "loss": 0.0061, "step": 17540 }, { "epoch": 5.44, "learning_rate": 2.986997399479896e-06, "loss": 0.0091, "step": 17541 }, { "epoch": 5.44, "learning_rate": 2.986597319463893e-06, "loss": 0.0049, "step": 17542 }, { "epoch": 5.44, "learning_rate": 2.98619723944789e-06, "loss": 0.0177, "step": 17543 }, { "epoch": 5.45, "learning_rate": 2.9857971594318867e-06, "loss": 0.0068, "step": 17544 }, { "epoch": 5.45, "learning_rate": 2.9853970794158834e-06, "loss": 0.0055, "step": 17545 }, { "epoch": 5.45, "learning_rate": 2.98499699939988e-06, "loss": 0.0032, "step": 17546 }, { "epoch": 5.45, "learning_rate": 2.984596919383877e-06, "loss": 0.0478, "step": 17547 }, { "epoch": 5.45, "learning_rate": 2.984196839367874e-06, "loss": 0.0013, "step": 17548 }, { "epoch": 5.45, "learning_rate": 2.983796759351871e-06, "loss": 0.0029, "step": 17549 }, { "epoch": 5.45, "learning_rate": 2.9833966793358676e-06, "loss": 0.005, "step": 17550 }, { "epoch": 5.45, "learning_rate": 2.982996599319864e-06, "loss": 0.0034, "step": 17551 }, { "epoch": 5.45, "learning_rate": 2.982596519303861e-06, "loss": 0.0049, "step": 17552 }, { "epoch": 5.45, "learning_rate": 2.9821964392878577e-06, "loss": 0.0013, "step": 17553 }, { "epoch": 5.45, "learning_rate": 2.981796359271855e-06, "loss": 0.0093, "step": 17554 }, { "epoch": 5.45, "learning_rate": 2.9813962792558515e-06, "loss": 0.0066, "step": 17555 }, { "epoch": 5.45, "learning_rate": 2.9809961992398486e-06, "loss": 0.0101, "step": 17556 }, { "epoch": 5.45, "learning_rate": 2.9805961192238453e-06, "loss": 0.0065, "step": 17557 }, { "epoch": 5.45, "learning_rate": 2.9801960392078416e-06, "loss": 0.0041, "step": 17558 }, { "epoch": 5.45, "learning_rate": 2.9797959591918387e-06, "loss": 0.0082, "step": 17559 }, { "epoch": 5.45, "learning_rate": 2.9793958791758353e-06, "loss": 0.0017, "step": 17560 }, { "epoch": 5.45, "learning_rate": 2.9789957991598325e-06, "loss": 0.0044, "step": 17561 }, { "epoch": 5.45, "learning_rate": 2.978595719143829e-06, "loss": 0.0237, "step": 17562 }, { "epoch": 5.45, "learning_rate": 2.9781956391278263e-06, "loss": 0.0021, "step": 17563 }, { "epoch": 5.45, "learning_rate": 2.9777955591118225e-06, "loss": 0.0106, "step": 17564 }, { "epoch": 5.45, "learning_rate": 2.977395479095819e-06, "loss": 0.0024, "step": 17565 }, { "epoch": 5.45, "learning_rate": 2.9769953990798163e-06, "loss": 0.0029, "step": 17566 }, { "epoch": 5.45, "learning_rate": 2.976595319063813e-06, "loss": 0.0116, "step": 17567 }, { "epoch": 5.45, "learning_rate": 2.97619523904781e-06, "loss": 0.0045, "step": 17568 }, { "epoch": 5.45, "learning_rate": 2.975795159031807e-06, "loss": 0.0051, "step": 17569 }, { "epoch": 5.45, "learning_rate": 2.975395079015803e-06, "loss": 0.0033, "step": 17570 }, { "epoch": 5.45, "learning_rate": 2.9749949989998e-06, "loss": 0.0026, "step": 17571 }, { "epoch": 5.45, "learning_rate": 2.974594918983797e-06, "loss": 0.0059, "step": 17572 }, { "epoch": 5.45, "learning_rate": 2.974194838967794e-06, "loss": 0.0028, "step": 17573 }, { "epoch": 5.45, "learning_rate": 2.9737947589517907e-06, "loss": 0.0022, "step": 17574 }, { "epoch": 5.45, "learning_rate": 2.9733946789357878e-06, "loss": 0.0098, "step": 17575 }, { "epoch": 5.46, "learning_rate": 2.972994598919784e-06, "loss": 0.0168, "step": 17576 }, { "epoch": 5.46, "learning_rate": 2.9725945189037807e-06, "loss": 0.0092, "step": 17577 }, { "epoch": 5.46, "learning_rate": 2.972194438887778e-06, "loss": 0.0178, "step": 17578 }, { "epoch": 5.46, "learning_rate": 2.9717943588717745e-06, "loss": 0.0066, "step": 17579 }, { "epoch": 5.46, "learning_rate": 2.9713942788557716e-06, "loss": 0.0099, "step": 17580 }, { "epoch": 5.46, "learning_rate": 2.9709941988397683e-06, "loss": 0.0048, "step": 17581 }, { "epoch": 5.46, "learning_rate": 2.9705941188237654e-06, "loss": 0.0123, "step": 17582 }, { "epoch": 5.46, "learning_rate": 2.9701940388077617e-06, "loss": 0.0124, "step": 17583 }, { "epoch": 5.46, "learning_rate": 2.9697939587917584e-06, "loss": 0.043, "step": 17584 }, { "epoch": 5.46, "learning_rate": 2.9693938787757555e-06, "loss": 0.0134, "step": 17585 }, { "epoch": 5.46, "learning_rate": 2.968993798759752e-06, "loss": 0.0027, "step": 17586 }, { "epoch": 5.46, "learning_rate": 2.9685937187437493e-06, "loss": 0.0188, "step": 17587 }, { "epoch": 5.46, "learning_rate": 2.968193638727746e-06, "loss": 0.0071, "step": 17588 }, { "epoch": 5.46, "learning_rate": 2.9677935587117422e-06, "loss": 0.0091, "step": 17589 }, { "epoch": 5.46, "learning_rate": 2.9673934786957393e-06, "loss": 0.0041, "step": 17590 }, { "epoch": 5.46, "learning_rate": 2.966993398679736e-06, "loss": 0.0018, "step": 17591 }, { "epoch": 5.46, "learning_rate": 2.966593318663733e-06, "loss": 0.0089, "step": 17592 }, { "epoch": 5.46, "learning_rate": 2.96619323864773e-06, "loss": 0.003, "step": 17593 }, { "epoch": 5.46, "learning_rate": 2.965793158631727e-06, "loss": 0.0192, "step": 17594 }, { "epoch": 5.46, "learning_rate": 2.965393078615723e-06, "loss": 0.0134, "step": 17595 }, { "epoch": 5.46, "learning_rate": 2.96499299859972e-06, "loss": 0.0049, "step": 17596 }, { "epoch": 5.46, "learning_rate": 2.964592918583717e-06, "loss": 0.011, "step": 17597 }, { "epoch": 5.46, "learning_rate": 2.9641928385677137e-06, "loss": 0.0025, "step": 17598 }, { "epoch": 5.46, "learning_rate": 2.9637927585517108e-06, "loss": 0.0044, "step": 17599 }, { "epoch": 5.46, "learning_rate": 2.9633926785357075e-06, "loss": 0.0227, "step": 17600 }, { "epoch": 5.46, "learning_rate": 2.9629925985197046e-06, "loss": 0.0059, "step": 17601 }, { "epoch": 5.46, "learning_rate": 2.962592518503701e-06, "loss": 0.009, "step": 17602 }, { "epoch": 5.46, "learning_rate": 2.9621924384876975e-06, "loss": 0.0048, "step": 17603 }, { "epoch": 5.46, "learning_rate": 2.9617923584716946e-06, "loss": 0.0025, "step": 17604 }, { "epoch": 5.46, "learning_rate": 2.9613922784556913e-06, "loss": 0.002, "step": 17605 }, { "epoch": 5.46, "learning_rate": 2.9609921984396884e-06, "loss": 0.0442, "step": 17606 }, { "epoch": 5.46, "learning_rate": 2.960592118423685e-06, "loss": 0.0098, "step": 17607 }, { "epoch": 5.47, "learning_rate": 2.9601920384076814e-06, "loss": 0.0053, "step": 17608 }, { "epoch": 5.47, "learning_rate": 2.9597919583916785e-06, "loss": 0.0059, "step": 17609 }, { "epoch": 5.47, "learning_rate": 2.959391878375675e-06, "loss": 0.0029, "step": 17610 }, { "epoch": 5.47, "learning_rate": 2.9589917983596723e-06, "loss": 0.0314, "step": 17611 }, { "epoch": 5.47, "learning_rate": 2.958591718343669e-06, "loss": 0.0027, "step": 17612 }, { "epoch": 5.47, "learning_rate": 2.958191638327666e-06, "loss": 0.0041, "step": 17613 }, { "epoch": 5.47, "learning_rate": 2.9577915583116623e-06, "loss": 0.0021, "step": 17614 }, { "epoch": 5.47, "learning_rate": 2.957391478295659e-06, "loss": 0.0053, "step": 17615 }, { "epoch": 5.47, "learning_rate": 2.956991398279656e-06, "loss": 0.0026, "step": 17616 }, { "epoch": 5.47, "learning_rate": 2.956591318263653e-06, "loss": 0.0049, "step": 17617 }, { "epoch": 5.47, "learning_rate": 2.95619123824765e-06, "loss": 0.0043, "step": 17618 }, { "epoch": 5.47, "learning_rate": 2.9557911582316466e-06, "loss": 0.0075, "step": 17619 }, { "epoch": 5.47, "learning_rate": 2.9553910782156437e-06, "loss": 0.0046, "step": 17620 }, { "epoch": 5.47, "learning_rate": 2.95499099819964e-06, "loss": 0.0166, "step": 17621 }, { "epoch": 5.47, "learning_rate": 2.9545909181836367e-06, "loss": 0.0365, "step": 17622 }, { "epoch": 5.47, "learning_rate": 2.9541908381676338e-06, "loss": 0.0036, "step": 17623 }, { "epoch": 5.47, "learning_rate": 2.9537907581516305e-06, "loss": 0.0063, "step": 17624 }, { "epoch": 5.47, "learning_rate": 2.9533906781356276e-06, "loss": 0.0103, "step": 17625 }, { "epoch": 5.47, "learning_rate": 2.9529905981196243e-06, "loss": 0.0606, "step": 17626 }, { "epoch": 5.47, "learning_rate": 2.9525905181036205e-06, "loss": 0.0082, "step": 17627 }, { "epoch": 5.47, "learning_rate": 2.9521904380876176e-06, "loss": 0.0043, "step": 17628 }, { "epoch": 5.47, "learning_rate": 2.9517903580716143e-06, "loss": 0.0052, "step": 17629 }, { "epoch": 5.47, "learning_rate": 2.9513902780556114e-06, "loss": 0.0256, "step": 17630 }, { "epoch": 5.47, "learning_rate": 2.950990198039608e-06, "loss": 0.0309, "step": 17631 }, { "epoch": 5.47, "learning_rate": 2.9505901180236052e-06, "loss": 0.0493, "step": 17632 }, { "epoch": 5.47, "learning_rate": 2.9501900380076015e-06, "loss": 0.0029, "step": 17633 }, { "epoch": 5.47, "learning_rate": 2.9497899579915986e-06, "loss": 0.0022, "step": 17634 }, { "epoch": 5.47, "learning_rate": 2.9493898779755953e-06, "loss": 0.0036, "step": 17635 }, { "epoch": 5.47, "learning_rate": 2.948989797959592e-06, "loss": 0.0432, "step": 17636 }, { "epoch": 5.47, "learning_rate": 2.948589717943589e-06, "loss": 0.01, "step": 17637 }, { "epoch": 5.47, "learning_rate": 2.9481896379275858e-06, "loss": 0.0166, "step": 17638 }, { "epoch": 5.47, "learning_rate": 2.947789557911583e-06, "loss": 0.1281, "step": 17639 }, { "epoch": 5.48, "learning_rate": 2.947389477895579e-06, "loss": 0.0044, "step": 17640 }, { "epoch": 5.48, "learning_rate": 2.9469893978795763e-06, "loss": 0.0107, "step": 17641 }, { "epoch": 5.48, "learning_rate": 2.946589317863573e-06, "loss": 0.0029, "step": 17642 }, { "epoch": 5.48, "learning_rate": 2.9461892378475696e-06, "loss": 0.0087, "step": 17643 }, { "epoch": 5.48, "learning_rate": 2.9457891578315667e-06, "loss": 0.0074, "step": 17644 }, { "epoch": 5.48, "learning_rate": 2.9453890778155634e-06, "loss": 0.0045, "step": 17645 }, { "epoch": 5.48, "learning_rate": 2.94498899779956e-06, "loss": 0.002, "step": 17646 }, { "epoch": 5.48, "learning_rate": 2.944588917783557e-06, "loss": 0.0025, "step": 17647 }, { "epoch": 5.48, "learning_rate": 2.944188837767554e-06, "loss": 0.0055, "step": 17648 }, { "epoch": 5.48, "learning_rate": 2.9437887577515506e-06, "loss": 0.0128, "step": 17649 }, { "epoch": 5.48, "learning_rate": 2.9433886777355473e-06, "loss": 0.0032, "step": 17650 }, { "epoch": 5.48, "learning_rate": 2.9429885977195444e-06, "loss": 0.0059, "step": 17651 }, { "epoch": 5.48, "learning_rate": 2.9425885177035407e-06, "loss": 0.0023, "step": 17652 }, { "epoch": 5.48, "learning_rate": 2.9421884376875378e-06, "loss": 0.0135, "step": 17653 }, { "epoch": 5.48, "learning_rate": 2.9417883576715344e-06, "loss": 0.0072, "step": 17654 }, { "epoch": 5.48, "learning_rate": 2.9413882776555316e-06, "loss": 0.0105, "step": 17655 }, { "epoch": 5.48, "learning_rate": 2.9409881976395282e-06, "loss": 0.0024, "step": 17656 }, { "epoch": 5.48, "learning_rate": 2.940588117623525e-06, "loss": 0.0198, "step": 17657 }, { "epoch": 5.48, "learning_rate": 2.9401880376075216e-06, "loss": 0.0165, "step": 17658 }, { "epoch": 5.48, "learning_rate": 2.9397879575915183e-06, "loss": 0.0293, "step": 17659 }, { "epoch": 5.48, "learning_rate": 2.9393878775755154e-06, "loss": 0.0684, "step": 17660 }, { "epoch": 5.48, "learning_rate": 2.938987797559512e-06, "loss": 0.0062, "step": 17661 }, { "epoch": 5.48, "learning_rate": 2.938587717543509e-06, "loss": 0.0038, "step": 17662 }, { "epoch": 5.48, "learning_rate": 2.938187637527506e-06, "loss": 0.0069, "step": 17663 }, { "epoch": 5.48, "learning_rate": 2.937787557511503e-06, "loss": 0.0189, "step": 17664 }, { "epoch": 5.48, "learning_rate": 2.9373874774954993e-06, "loss": 0.1063, "step": 17665 }, { "epoch": 5.48, "learning_rate": 2.936987397479496e-06, "loss": 0.0304, "step": 17666 }, { "epoch": 5.48, "learning_rate": 2.936587317463493e-06, "loss": 0.0105, "step": 17667 }, { "epoch": 5.48, "learning_rate": 2.9361872374474897e-06, "loss": 0.0042, "step": 17668 }, { "epoch": 5.48, "learning_rate": 2.935787157431487e-06, "loss": 0.0141, "step": 17669 }, { "epoch": 5.48, "learning_rate": 2.9353870774154835e-06, "loss": 0.0115, "step": 17670 }, { "epoch": 5.48, "learning_rate": 2.93498699739948e-06, "loss": 0.0034, "step": 17671 }, { "epoch": 5.49, "learning_rate": 2.934586917383477e-06, "loss": 0.0026, "step": 17672 }, { "epoch": 5.49, "learning_rate": 2.9341868373674736e-06, "loss": 0.0101, "step": 17673 }, { "epoch": 5.49, "learning_rate": 2.9337867573514707e-06, "loss": 0.0047, "step": 17674 }, { "epoch": 5.49, "learning_rate": 2.9333866773354674e-06, "loss": 0.0053, "step": 17675 }, { "epoch": 5.49, "learning_rate": 2.9329865973194645e-06, "loss": 0.0025, "step": 17676 }, { "epoch": 5.49, "learning_rate": 2.9325865173034608e-06, "loss": 0.0045, "step": 17677 }, { "epoch": 5.49, "learning_rate": 2.9321864372874575e-06, "loss": 0.0037, "step": 17678 }, { "epoch": 5.49, "learning_rate": 2.9317863572714546e-06, "loss": 0.0062, "step": 17679 }, { "epoch": 5.49, "learning_rate": 2.9313862772554513e-06, "loss": 0.0129, "step": 17680 }, { "epoch": 5.49, "learning_rate": 2.9309861972394484e-06, "loss": 0.0039, "step": 17681 }, { "epoch": 5.49, "learning_rate": 2.930586117223445e-06, "loss": 0.0023, "step": 17682 }, { "epoch": 5.49, "learning_rate": 2.930186037207442e-06, "loss": 0.0076, "step": 17683 }, { "epoch": 5.49, "learning_rate": 2.9297859571914384e-06, "loss": 0.0059, "step": 17684 }, { "epoch": 5.49, "learning_rate": 2.929385877175435e-06, "loss": 0.0033, "step": 17685 }, { "epoch": 5.49, "learning_rate": 2.9289857971594322e-06, "loss": 0.0329, "step": 17686 }, { "epoch": 5.49, "learning_rate": 2.928585717143429e-06, "loss": 0.0105, "step": 17687 }, { "epoch": 5.49, "learning_rate": 2.928185637127426e-06, "loss": 0.0028, "step": 17688 }, { "epoch": 5.49, "learning_rate": 2.9277855571114227e-06, "loss": 0.0114, "step": 17689 }, { "epoch": 5.49, "learning_rate": 2.927385477095419e-06, "loss": 0.0044, "step": 17690 }, { "epoch": 5.49, "learning_rate": 2.926985397079416e-06, "loss": 0.0112, "step": 17691 }, { "epoch": 5.49, "learning_rate": 2.9265853170634128e-06, "loss": 0.0039, "step": 17692 }, { "epoch": 5.49, "learning_rate": 2.92618523704741e-06, "loss": 0.0053, "step": 17693 }, { "epoch": 5.49, "learning_rate": 2.9257851570314066e-06, "loss": 0.0389, "step": 17694 }, { "epoch": 5.49, "learning_rate": 2.9253850770154037e-06, "loss": 0.0083, "step": 17695 }, { "epoch": 5.49, "learning_rate": 2.9249849969994e-06, "loss": 0.005, "step": 17696 }, { "epoch": 5.49, "learning_rate": 2.9245849169833966e-06, "loss": 0.0197, "step": 17697 }, { "epoch": 5.49, "learning_rate": 2.9241848369673937e-06, "loss": 0.0048, "step": 17698 }, { "epoch": 5.49, "learning_rate": 2.9237847569513904e-06, "loss": 0.0041, "step": 17699 }, { "epoch": 5.49, "learning_rate": 2.9233846769353875e-06, "loss": 0.0044, "step": 17700 }, { "epoch": 5.49, "learning_rate": 2.922984596919384e-06, "loss": 0.0279, "step": 17701 }, { "epoch": 5.49, "learning_rate": 2.9225845169033813e-06, "loss": 0.006, "step": 17702 }, { "epoch": 5.49, "learning_rate": 2.9221844368873776e-06, "loss": 0.008, "step": 17703 }, { "epoch": 5.49, "learning_rate": 2.9217843568713743e-06, "loss": 0.0081, "step": 17704 }, { "epoch": 5.5, "learning_rate": 2.9213842768553714e-06, "loss": 0.008, "step": 17705 }, { "epoch": 5.5, "learning_rate": 2.920984196839368e-06, "loss": 0.002, "step": 17706 }, { "epoch": 5.5, "learning_rate": 2.920584116823365e-06, "loss": 0.0056, "step": 17707 }, { "epoch": 5.5, "learning_rate": 2.920184036807362e-06, "loss": 0.004, "step": 17708 }, { "epoch": 5.5, "learning_rate": 2.919783956791358e-06, "loss": 0.0226, "step": 17709 }, { "epoch": 5.5, "learning_rate": 2.9193838767753552e-06, "loss": 0.0017, "step": 17710 }, { "epoch": 5.5, "learning_rate": 2.918983796759352e-06, "loss": 0.0133, "step": 17711 }, { "epoch": 5.5, "learning_rate": 2.918583716743349e-06, "loss": 0.0037, "step": 17712 }, { "epoch": 5.5, "learning_rate": 2.9181836367273457e-06, "loss": 0.0043, "step": 17713 }, { "epoch": 5.5, "learning_rate": 2.917783556711343e-06, "loss": 0.0046, "step": 17714 }, { "epoch": 5.5, "learning_rate": 2.917383476695339e-06, "loss": 0.0044, "step": 17715 }, { "epoch": 5.5, "learning_rate": 2.9169833966793358e-06, "loss": 0.0073, "step": 17716 }, { "epoch": 5.5, "learning_rate": 2.916583316663333e-06, "loss": 0.0042, "step": 17717 }, { "epoch": 5.5, "learning_rate": 2.9161832366473296e-06, "loss": 0.0016, "step": 17718 }, { "epoch": 5.5, "learning_rate": 2.9157831566313267e-06, "loss": 0.0028, "step": 17719 }, { "epoch": 5.5, "learning_rate": 2.9153830766153234e-06, "loss": 0.0079, "step": 17720 }, { "epoch": 5.5, "learning_rate": 2.9149829965993196e-06, "loss": 0.0034, "step": 17721 }, { "epoch": 5.5, "learning_rate": 2.9145829165833167e-06, "loss": 0.005, "step": 17722 }, { "epoch": 5.5, "learning_rate": 2.9141828365673134e-06, "loss": 0.0091, "step": 17723 }, { "epoch": 5.5, "learning_rate": 2.9137827565513105e-06, "loss": 0.0206, "step": 17724 }, { "epoch": 5.5, "learning_rate": 2.9133826765353072e-06, "loss": 0.0033, "step": 17725 }, { "epoch": 5.5, "learning_rate": 2.9129825965193043e-06, "loss": 0.0865, "step": 17726 }, { "epoch": 5.5, "learning_rate": 2.912582516503301e-06, "loss": 0.0055, "step": 17727 }, { "epoch": 5.5, "learning_rate": 2.9121824364872973e-06, "loss": 0.0033, "step": 17728 }, { "epoch": 5.5, "learning_rate": 2.9117823564712944e-06, "loss": 0.006, "step": 17729 }, { "epoch": 5.5, "learning_rate": 2.911382276455291e-06, "loss": 0.0308, "step": 17730 }, { "epoch": 5.5, "learning_rate": 2.910982196439288e-06, "loss": 0.0728, "step": 17731 }, { "epoch": 5.5, "learning_rate": 2.910582116423285e-06, "loss": 0.0173, "step": 17732 }, { "epoch": 5.5, "learning_rate": 2.910182036407282e-06, "loss": 0.0089, "step": 17733 }, { "epoch": 5.5, "learning_rate": 2.9097819563912782e-06, "loss": 0.008, "step": 17734 }, { "epoch": 5.5, "learning_rate": 2.909381876375275e-06, "loss": 0.0199, "step": 17735 }, { "epoch": 5.5, "learning_rate": 2.908981796359272e-06, "loss": 0.0038, "step": 17736 }, { "epoch": 5.51, "learning_rate": 2.9085817163432687e-06, "loss": 0.0017, "step": 17737 }, { "epoch": 5.51, "learning_rate": 2.908181636327266e-06, "loss": 0.0059, "step": 17738 }, { "epoch": 5.51, "learning_rate": 2.9077815563112625e-06, "loss": 0.0039, "step": 17739 }, { "epoch": 5.51, "learning_rate": 2.907381476295259e-06, "loss": 0.0028, "step": 17740 }, { "epoch": 5.51, "learning_rate": 2.906981396279256e-06, "loss": 0.0053, "step": 17741 }, { "epoch": 5.51, "learning_rate": 2.906581316263253e-06, "loss": 0.0033, "step": 17742 }, { "epoch": 5.51, "learning_rate": 2.9061812362472497e-06, "loss": 0.0044, "step": 17743 }, { "epoch": 5.51, "learning_rate": 2.9057811562312464e-06, "loss": 0.0165, "step": 17744 }, { "epoch": 5.51, "learning_rate": 2.9053810762152435e-06, "loss": 0.0063, "step": 17745 }, { "epoch": 5.51, "learning_rate": 2.90498099619924e-06, "loss": 0.0063, "step": 17746 }, { "epoch": 5.51, "learning_rate": 2.904580916183237e-06, "loss": 0.0046, "step": 17747 }, { "epoch": 5.51, "learning_rate": 2.9041808361672335e-06, "loss": 0.0032, "step": 17748 }, { "epoch": 5.51, "learning_rate": 2.9037807561512307e-06, "loss": 0.004, "step": 17749 }, { "epoch": 5.51, "learning_rate": 2.9033806761352273e-06, "loss": 0.0135, "step": 17750 }, { "epoch": 5.51, "learning_rate": 2.902980596119224e-06, "loss": 0.0036, "step": 17751 }, { "epoch": 5.51, "learning_rate": 2.902580516103221e-06, "loss": 0.0792, "step": 17752 }, { "epoch": 5.51, "learning_rate": 2.9021804360872174e-06, "loss": 0.004, "step": 17753 }, { "epoch": 5.51, "learning_rate": 2.9017803560712145e-06, "loss": 0.0019, "step": 17754 }, { "epoch": 5.51, "learning_rate": 2.901380276055211e-06, "loss": 0.0096, "step": 17755 }, { "epoch": 5.51, "learning_rate": 2.9009801960392083e-06, "loss": 0.0548, "step": 17756 }, { "epoch": 5.51, "learning_rate": 2.900580116023205e-06, "loss": 0.0324, "step": 17757 }, { "epoch": 5.51, "learning_rate": 2.9001800360072017e-06, "loss": 0.0065, "step": 17758 }, { "epoch": 5.51, "learning_rate": 2.8997799559911984e-06, "loss": 0.011, "step": 17759 }, { "epoch": 5.51, "learning_rate": 2.899379875975195e-06, "loss": 0.003, "step": 17760 }, { "epoch": 5.51, "learning_rate": 2.898979795959192e-06, "loss": 0.0035, "step": 17761 }, { "epoch": 5.51, "learning_rate": 2.898579715943189e-06, "loss": 0.0055, "step": 17762 }, { "epoch": 5.51, "learning_rate": 2.898179635927186e-06, "loss": 0.0063, "step": 17763 }, { "epoch": 5.51, "learning_rate": 2.8977795559111826e-06, "loss": 0.0042, "step": 17764 }, { "epoch": 5.51, "learning_rate": 2.8973794758951793e-06, "loss": 0.0054, "step": 17765 }, { "epoch": 5.51, "learning_rate": 2.896979395879176e-06, "loss": 0.0043, "step": 17766 }, { "epoch": 5.51, "learning_rate": 2.8965793158631727e-06, "loss": 0.0354, "step": 17767 }, { "epoch": 5.51, "learning_rate": 2.89617923584717e-06, "loss": 0.0041, "step": 17768 }, { "epoch": 5.52, "learning_rate": 2.8957791558311665e-06, "loss": 0.0033, "step": 17769 }, { "epoch": 5.52, "learning_rate": 2.8953790758151636e-06, "loss": 0.0025, "step": 17770 }, { "epoch": 5.52, "learning_rate": 2.8949789957991603e-06, "loss": 0.0344, "step": 17771 }, { "epoch": 5.52, "learning_rate": 2.8945789157831566e-06, "loss": 0.0029, "step": 17772 }, { "epoch": 5.52, "learning_rate": 2.8941788357671537e-06, "loss": 0.0069, "step": 17773 }, { "epoch": 5.52, "learning_rate": 2.8937787557511504e-06, "loss": 0.0205, "step": 17774 }, { "epoch": 5.52, "learning_rate": 2.8933786757351475e-06, "loss": 0.0056, "step": 17775 }, { "epoch": 5.52, "learning_rate": 2.892978595719144e-06, "loss": 0.0135, "step": 17776 }, { "epoch": 5.52, "learning_rate": 2.8925785157031413e-06, "loss": 0.0038, "step": 17777 }, { "epoch": 5.52, "learning_rate": 2.8921784356871375e-06, "loss": 0.018, "step": 17778 }, { "epoch": 5.52, "learning_rate": 2.891778355671134e-06, "loss": 0.0065, "step": 17779 }, { "epoch": 5.52, "learning_rate": 2.8913782756551313e-06, "loss": 0.0093, "step": 17780 }, { "epoch": 5.52, "learning_rate": 2.890978195639128e-06, "loss": 0.0869, "step": 17781 }, { "epoch": 5.52, "learning_rate": 2.890578115623125e-06, "loss": 0.0077, "step": 17782 }, { "epoch": 5.52, "learning_rate": 2.890178035607122e-06, "loss": 0.0051, "step": 17783 }, { "epoch": 5.52, "learning_rate": 2.889777955591119e-06, "loss": 0.01, "step": 17784 }, { "epoch": 5.52, "learning_rate": 2.889377875575115e-06, "loss": 0.004, "step": 17785 }, { "epoch": 5.52, "learning_rate": 2.888977795559112e-06, "loss": 0.0057, "step": 17786 }, { "epoch": 5.52, "learning_rate": 2.888577715543109e-06, "loss": 0.0104, "step": 17787 }, { "epoch": 5.52, "learning_rate": 2.8881776355271057e-06, "loss": 0.0055, "step": 17788 }, { "epoch": 5.52, "learning_rate": 2.8877775555111028e-06, "loss": 0.0699, "step": 17789 }, { "epoch": 5.52, "learning_rate": 2.8873774754950995e-06, "loss": 0.0096, "step": 17790 }, { "epoch": 5.52, "learning_rate": 2.8869773954790957e-06, "loss": 0.0019, "step": 17791 }, { "epoch": 5.52, "learning_rate": 2.886577315463093e-06, "loss": 0.0554, "step": 17792 }, { "epoch": 5.52, "learning_rate": 2.8861772354470895e-06, "loss": 0.0028, "step": 17793 }, { "epoch": 5.52, "learning_rate": 2.8857771554310866e-06, "loss": 0.0183, "step": 17794 }, { "epoch": 5.52, "learning_rate": 2.8853770754150833e-06, "loss": 0.0021, "step": 17795 }, { "epoch": 5.52, "learning_rate": 2.8849769953990804e-06, "loss": 0.0044, "step": 17796 }, { "epoch": 5.52, "learning_rate": 2.8845769153830767e-06, "loss": 0.0096, "step": 17797 }, { "epoch": 5.52, "learning_rate": 2.8841768353670734e-06, "loss": 0.0083, "step": 17798 }, { "epoch": 5.52, "learning_rate": 2.8837767553510705e-06, "loss": 0.0033, "step": 17799 }, { "epoch": 5.52, "learning_rate": 2.883376675335067e-06, "loss": 0.0041, "step": 17800 }, { "epoch": 5.53, "learning_rate": 2.8829765953190643e-06, "loss": 0.0071, "step": 17801 }, { "epoch": 5.53, "learning_rate": 2.882576515303061e-06, "loss": 0.0188, "step": 17802 }, { "epoch": 5.53, "learning_rate": 2.8821764352870572e-06, "loss": 0.017, "step": 17803 }, { "epoch": 5.53, "learning_rate": 2.8817763552710543e-06, "loss": 0.0086, "step": 17804 }, { "epoch": 5.53, "learning_rate": 2.881376275255051e-06, "loss": 0.0117, "step": 17805 }, { "epoch": 5.53, "learning_rate": 2.880976195239048e-06, "loss": 0.0049, "step": 17806 }, { "epoch": 5.53, "learning_rate": 2.880576115223045e-06, "loss": 0.0375, "step": 17807 }, { "epoch": 5.53, "learning_rate": 2.880176035207042e-06, "loss": 0.0031, "step": 17808 }, { "epoch": 5.53, "learning_rate": 2.8797759551910386e-06, "loss": 0.0058, "step": 17809 }, { "epoch": 5.53, "learning_rate": 2.879375875175035e-06, "loss": 0.0099, "step": 17810 }, { "epoch": 5.53, "learning_rate": 2.878975795159032e-06, "loss": 0.0134, "step": 17811 }, { "epoch": 5.53, "learning_rate": 2.8785757151430287e-06, "loss": 0.0094, "step": 17812 }, { "epoch": 5.53, "learning_rate": 2.8781756351270258e-06, "loss": 0.0068, "step": 17813 }, { "epoch": 5.53, "learning_rate": 2.8777755551110225e-06, "loss": 0.0013, "step": 17814 }, { "epoch": 5.53, "learning_rate": 2.8773754750950196e-06, "loss": 0.0087, "step": 17815 }, { "epoch": 5.53, "learning_rate": 2.876975395079016e-06, "loss": 0.0107, "step": 17816 }, { "epoch": 5.53, "learning_rate": 2.8765753150630125e-06, "loss": 0.0076, "step": 17817 }, { "epoch": 5.53, "learning_rate": 2.8761752350470096e-06, "loss": 0.0565, "step": 17818 }, { "epoch": 5.53, "learning_rate": 2.8757751550310063e-06, "loss": 0.0086, "step": 17819 }, { "epoch": 5.53, "learning_rate": 2.8753750750150034e-06, "loss": 0.0137, "step": 17820 }, { "epoch": 5.53, "learning_rate": 2.874974994999e-06, "loss": 0.0039, "step": 17821 }, { "epoch": 5.53, "learning_rate": 2.8745749149829964e-06, "loss": 0.0097, "step": 17822 }, { "epoch": 5.53, "learning_rate": 2.8741748349669935e-06, "loss": 0.0032, "step": 17823 }, { "epoch": 5.53, "learning_rate": 2.87377475495099e-06, "loss": 0.0026, "step": 17824 }, { "epoch": 5.53, "learning_rate": 2.8733746749349873e-06, "loss": 0.0051, "step": 17825 }, { "epoch": 5.53, "learning_rate": 2.872974594918984e-06, "loss": 0.0033, "step": 17826 }, { "epoch": 5.53, "learning_rate": 2.872574514902981e-06, "loss": 0.0074, "step": 17827 }, { "epoch": 5.53, "learning_rate": 2.8721744348869778e-06, "loss": 0.0679, "step": 17828 }, { "epoch": 5.53, "learning_rate": 2.871774354870974e-06, "loss": 0.0093, "step": 17829 }, { "epoch": 5.53, "learning_rate": 2.871374274854971e-06, "loss": 0.0113, "step": 17830 }, { "epoch": 5.53, "learning_rate": 2.870974194838968e-06, "loss": 0.0103, "step": 17831 }, { "epoch": 5.53, "learning_rate": 2.870574114822965e-06, "loss": 0.0037, "step": 17832 }, { "epoch": 5.53, "learning_rate": 2.8701740348069616e-06, "loss": 0.003, "step": 17833 }, { "epoch": 5.54, "learning_rate": 2.8697739547909587e-06, "loss": 0.0059, "step": 17834 }, { "epoch": 5.54, "learning_rate": 2.869373874774955e-06, "loss": 0.0035, "step": 17835 }, { "epoch": 5.54, "learning_rate": 2.8689737947589517e-06, "loss": 0.0138, "step": 17836 }, { "epoch": 5.54, "learning_rate": 2.8685737147429488e-06, "loss": 0.0535, "step": 17837 }, { "epoch": 5.54, "learning_rate": 2.8681736347269455e-06, "loss": 0.0324, "step": 17838 }, { "epoch": 5.54, "learning_rate": 2.8677735547109426e-06, "loss": 0.015, "step": 17839 }, { "epoch": 5.54, "learning_rate": 2.8673734746949393e-06, "loss": 0.0527, "step": 17840 }, { "epoch": 5.54, "learning_rate": 2.866973394678936e-06, "loss": 0.0047, "step": 17841 }, { "epoch": 5.54, "learning_rate": 2.8665733146629326e-06, "loss": 0.0061, "step": 17842 }, { "epoch": 5.54, "learning_rate": 2.8661732346469293e-06, "loss": 0.0079, "step": 17843 }, { "epoch": 5.54, "learning_rate": 2.8657731546309264e-06, "loss": 0.0029, "step": 17844 }, { "epoch": 5.54, "learning_rate": 2.865373074614923e-06, "loss": 0.0108, "step": 17845 }, { "epoch": 5.54, "learning_rate": 2.8649729945989202e-06, "loss": 0.008, "step": 17846 }, { "epoch": 5.54, "learning_rate": 2.864572914582917e-06, "loss": 0.021, "step": 17847 }, { "epoch": 5.54, "learning_rate": 2.8641728345669136e-06, "loss": 0.0066, "step": 17848 }, { "epoch": 5.54, "learning_rate": 2.8637727545509103e-06, "loss": 0.0103, "step": 17849 }, { "epoch": 5.54, "learning_rate": 2.8633726745349074e-06, "loss": 0.0072, "step": 17850 }, { "epoch": 5.54, "learning_rate": 2.862972594518904e-06, "loss": 0.009, "step": 17851 }, { "epoch": 5.54, "learning_rate": 2.8625725145029008e-06, "loss": 0.0147, "step": 17852 }, { "epoch": 5.54, "learning_rate": 2.862172434486898e-06, "loss": 0.0064, "step": 17853 }, { "epoch": 5.54, "learning_rate": 2.861772354470894e-06, "loss": 0.0037, "step": 17854 }, { "epoch": 5.54, "learning_rate": 2.8613722744548913e-06, "loss": 0.0063, "step": 17855 }, { "epoch": 5.54, "learning_rate": 2.860972194438888e-06, "loss": 0.0031, "step": 17856 }, { "epoch": 5.54, "learning_rate": 2.860572114422885e-06, "loss": 0.002, "step": 17857 }, { "epoch": 5.54, "learning_rate": 2.8601720344068817e-06, "loss": 0.0022, "step": 17858 }, { "epoch": 5.54, "learning_rate": 2.8597719543908784e-06, "loss": 0.0212, "step": 17859 }, { "epoch": 5.54, "learning_rate": 2.859371874374875e-06, "loss": 0.0155, "step": 17860 }, { "epoch": 5.54, "learning_rate": 2.858971794358872e-06, "loss": 0.0104, "step": 17861 }, { "epoch": 5.54, "learning_rate": 2.858571714342869e-06, "loss": 0.0035, "step": 17862 }, { "epoch": 5.54, "learning_rate": 2.8581716343268656e-06, "loss": 0.0062, "step": 17863 }, { "epoch": 5.54, "learning_rate": 2.8577715543108627e-06, "loss": 0.0052, "step": 17864 }, { "epoch": 5.54, "learning_rate": 2.8573714742948594e-06, "loss": 0.0035, "step": 17865 }, { "epoch": 5.55, "learning_rate": 2.856971394278856e-06, "loss": 0.0066, "step": 17866 }, { "epoch": 5.55, "learning_rate": 2.8565713142628528e-06, "loss": 0.0105, "step": 17867 }, { "epoch": 5.55, "learning_rate": 2.8561712342468494e-06, "loss": 0.0018, "step": 17868 }, { "epoch": 5.55, "learning_rate": 2.8557711542308466e-06, "loss": 0.0033, "step": 17869 }, { "epoch": 5.55, "learning_rate": 2.8553710742148432e-06, "loss": 0.0086, "step": 17870 }, { "epoch": 5.55, "learning_rate": 2.8549709941988404e-06, "loss": 0.0059, "step": 17871 }, { "epoch": 5.55, "learning_rate": 2.854570914182837e-06, "loss": 0.0173, "step": 17872 }, { "epoch": 5.55, "learning_rate": 2.8541708341668333e-06, "loss": 0.0035, "step": 17873 }, { "epoch": 5.55, "learning_rate": 2.8537707541508304e-06, "loss": 0.0077, "step": 17874 }, { "epoch": 5.55, "learning_rate": 2.853370674134827e-06, "loss": 0.0208, "step": 17875 }, { "epoch": 5.55, "learning_rate": 2.852970594118824e-06, "loss": 0.0126, "step": 17876 }, { "epoch": 5.55, "learning_rate": 2.852570514102821e-06, "loss": 0.0087, "step": 17877 }, { "epoch": 5.55, "learning_rate": 2.852170434086818e-06, "loss": 0.0043, "step": 17878 }, { "epoch": 5.55, "learning_rate": 2.8517703540708143e-06, "loss": 0.019, "step": 17879 }, { "epoch": 5.55, "learning_rate": 2.851370274054811e-06, "loss": 0.0078, "step": 17880 }, { "epoch": 5.55, "learning_rate": 2.850970194038808e-06, "loss": 0.0051, "step": 17881 }, { "epoch": 5.55, "learning_rate": 2.8505701140228048e-06, "loss": 0.004, "step": 17882 }, { "epoch": 5.55, "learning_rate": 2.850170034006802e-06, "loss": 0.0029, "step": 17883 }, { "epoch": 5.55, "learning_rate": 2.8497699539907985e-06, "loss": 0.0042, "step": 17884 }, { "epoch": 5.55, "learning_rate": 2.849369873974795e-06, "loss": 0.0039, "step": 17885 }, { "epoch": 5.55, "learning_rate": 2.848969793958792e-06, "loss": 0.0131, "step": 17886 }, { "epoch": 5.55, "learning_rate": 2.8485697139427886e-06, "loss": 0.0348, "step": 17887 }, { "epoch": 5.55, "learning_rate": 2.8481696339267857e-06, "loss": 0.0279, "step": 17888 }, { "epoch": 5.55, "learning_rate": 2.8477695539107824e-06, "loss": 0.0038, "step": 17889 }, { "epoch": 5.55, "learning_rate": 2.8473694738947795e-06, "loss": 0.0035, "step": 17890 }, { "epoch": 5.55, "learning_rate": 2.846969393878776e-06, "loss": 0.0555, "step": 17891 }, { "epoch": 5.55, "learning_rate": 2.8465693138627725e-06, "loss": 0.0034, "step": 17892 }, { "epoch": 5.55, "learning_rate": 2.8461692338467696e-06, "loss": 0.0041, "step": 17893 }, { "epoch": 5.55, "learning_rate": 2.8457691538307663e-06, "loss": 0.0055, "step": 17894 }, { "epoch": 5.55, "learning_rate": 2.8453690738147634e-06, "loss": 0.013, "step": 17895 }, { "epoch": 5.55, "learning_rate": 2.84496899379876e-06, "loss": 0.025, "step": 17896 }, { "epoch": 5.55, "learning_rate": 2.844568913782757e-06, "loss": 0.0105, "step": 17897 }, { "epoch": 5.56, "learning_rate": 2.8441688337667534e-06, "loss": 0.0024, "step": 17898 }, { "epoch": 5.56, "learning_rate": 2.84376875375075e-06, "loss": 0.0129, "step": 17899 }, { "epoch": 5.56, "learning_rate": 2.8433686737347472e-06, "loss": 0.0041, "step": 17900 }, { "epoch": 5.56, "learning_rate": 2.842968593718744e-06, "loss": 0.0036, "step": 17901 }, { "epoch": 5.56, "learning_rate": 2.842568513702741e-06, "loss": 0.0099, "step": 17902 }, { "epoch": 5.56, "learning_rate": 2.8421684336867377e-06, "loss": 0.0214, "step": 17903 }, { "epoch": 5.56, "learning_rate": 2.841768353670734e-06, "loss": 0.0044, "step": 17904 }, { "epoch": 5.56, "learning_rate": 2.841368273654731e-06, "loss": 0.0162, "step": 17905 }, { "epoch": 5.56, "learning_rate": 2.8409681936387278e-06, "loss": 0.0027, "step": 17906 }, { "epoch": 5.56, "learning_rate": 2.840568113622725e-06, "loss": 0.0095, "step": 17907 }, { "epoch": 5.56, "learning_rate": 2.8401680336067216e-06, "loss": 0.0071, "step": 17908 }, { "epoch": 5.56, "learning_rate": 2.8397679535907187e-06, "loss": 0.0099, "step": 17909 }, { "epoch": 5.56, "learning_rate": 2.8393678735747154e-06, "loss": 0.0083, "step": 17910 }, { "epoch": 5.56, "learning_rate": 2.8389677935587116e-06, "loss": 0.0029, "step": 17911 }, { "epoch": 5.56, "learning_rate": 2.8385677135427087e-06, "loss": 0.0051, "step": 17912 }, { "epoch": 5.56, "learning_rate": 2.8381676335267054e-06, "loss": 0.0033, "step": 17913 }, { "epoch": 5.56, "learning_rate": 2.8377675535107025e-06, "loss": 0.0041, "step": 17914 }, { "epoch": 5.56, "learning_rate": 2.837367473494699e-06, "loss": 0.0078, "step": 17915 }, { "epoch": 5.56, "learning_rate": 2.8369673934786963e-06, "loss": 0.0256, "step": 17916 }, { "epoch": 5.56, "learning_rate": 2.8365673134626926e-06, "loss": 0.0042, "step": 17917 }, { "epoch": 5.56, "learning_rate": 2.8361672334466893e-06, "loss": 0.0166, "step": 17918 }, { "epoch": 5.56, "learning_rate": 2.8357671534306864e-06, "loss": 0.0068, "step": 17919 }, { "epoch": 5.56, "learning_rate": 2.835367073414683e-06, "loss": 0.0016, "step": 17920 }, { "epoch": 5.56, "learning_rate": 2.83496699339868e-06, "loss": 0.0066, "step": 17921 }, { "epoch": 5.56, "learning_rate": 2.834566913382677e-06, "loss": 0.0132, "step": 17922 }, { "epoch": 5.56, "learning_rate": 2.834166833366673e-06, "loss": 0.003, "step": 17923 }, { "epoch": 5.56, "learning_rate": 2.8337667533506702e-06, "loss": 0.0041, "step": 17924 }, { "epoch": 5.56, "learning_rate": 2.833366673334667e-06, "loss": 0.0087, "step": 17925 }, { "epoch": 5.56, "learning_rate": 2.832966593318664e-06, "loss": 0.0212, "step": 17926 }, { "epoch": 5.56, "learning_rate": 2.8325665133026607e-06, "loss": 0.0236, "step": 17927 }, { "epoch": 5.56, "learning_rate": 2.832166433286658e-06, "loss": 0.0171, "step": 17928 }, { "epoch": 5.56, "learning_rate": 2.8317663532706545e-06, "loss": 0.0094, "step": 17929 }, { "epoch": 5.57, "learning_rate": 2.8313662732546508e-06, "loss": 0.0055, "step": 17930 }, { "epoch": 5.57, "learning_rate": 2.830966193238648e-06, "loss": 0.0105, "step": 17931 }, { "epoch": 5.57, "learning_rate": 2.8305661132226446e-06, "loss": 0.0439, "step": 17932 }, { "epoch": 5.57, "learning_rate": 2.8301660332066417e-06, "loss": 0.0095, "step": 17933 }, { "epoch": 5.57, "learning_rate": 2.8297659531906384e-06, "loss": 0.0019, "step": 17934 }, { "epoch": 5.57, "learning_rate": 2.8293658731746355e-06, "loss": 0.0129, "step": 17935 }, { "epoch": 5.57, "learning_rate": 2.8289657931586317e-06, "loss": 0.0055, "step": 17936 }, { "epoch": 5.57, "learning_rate": 2.8285657131426284e-06, "loss": 0.002, "step": 17937 }, { "epoch": 5.57, "learning_rate": 2.8281656331266255e-06, "loss": 0.0158, "step": 17938 }, { "epoch": 5.57, "learning_rate": 2.8277655531106222e-06, "loss": 0.01, "step": 17939 }, { "epoch": 5.57, "learning_rate": 2.8273654730946193e-06, "loss": 0.0029, "step": 17940 }, { "epoch": 5.57, "learning_rate": 2.826965393078616e-06, "loss": 0.0181, "step": 17941 }, { "epoch": 5.57, "learning_rate": 2.8265653130626127e-06, "loss": 0.0069, "step": 17942 }, { "epoch": 5.57, "learning_rate": 2.8261652330466094e-06, "loss": 0.0091, "step": 17943 }, { "epoch": 5.57, "learning_rate": 2.825765153030606e-06, "loss": 0.0021, "step": 17944 }, { "epoch": 5.57, "learning_rate": 2.825365073014603e-06, "loss": 0.0082, "step": 17945 }, { "epoch": 5.57, "learning_rate": 2.8249649929986e-06, "loss": 0.0016, "step": 17946 }, { "epoch": 5.57, "learning_rate": 2.824564912982597e-06, "loss": 0.0031, "step": 17947 }, { "epoch": 5.57, "learning_rate": 2.8241648329665932e-06, "loss": 0.0029, "step": 17948 }, { "epoch": 5.57, "learning_rate": 2.8237647529505904e-06, "loss": 0.0041, "step": 17949 }, { "epoch": 5.57, "learning_rate": 2.823364672934587e-06, "loss": 0.0044, "step": 17950 }, { "epoch": 5.57, "learning_rate": 2.8229645929185837e-06, "loss": 0.0089, "step": 17951 }, { "epoch": 5.57, "learning_rate": 2.822564512902581e-06, "loss": 0.007, "step": 17952 }, { "epoch": 5.57, "learning_rate": 2.8221644328865775e-06, "loss": 0.0021, "step": 17953 }, { "epoch": 5.57, "learning_rate": 2.8217643528705746e-06, "loss": 0.0218, "step": 17954 }, { "epoch": 5.57, "learning_rate": 2.821364272854571e-06, "loss": 0.0052, "step": 17955 }, { "epoch": 5.57, "learning_rate": 2.820964192838568e-06, "loss": 0.0089, "step": 17956 }, { "epoch": 5.57, "learning_rate": 2.8205641128225647e-06, "loss": 0.0123, "step": 17957 }, { "epoch": 5.57, "learning_rate": 2.820164032806562e-06, "loss": 0.0159, "step": 17958 }, { "epoch": 5.57, "learning_rate": 2.8197639527905585e-06, "loss": 0.0288, "step": 17959 }, { "epoch": 5.57, "learning_rate": 2.819363872774555e-06, "loss": 0.0052, "step": 17960 }, { "epoch": 5.57, "learning_rate": 2.818963792758552e-06, "loss": 0.0092, "step": 17961 }, { "epoch": 5.58, "learning_rate": 2.8185637127425485e-06, "loss": 0.0028, "step": 17962 }, { "epoch": 5.58, "learning_rate": 2.8181636327265457e-06, "loss": 0.0114, "step": 17963 }, { "epoch": 5.58, "learning_rate": 2.8177635527105423e-06, "loss": 0.0024, "step": 17964 }, { "epoch": 5.58, "learning_rate": 2.8173634726945395e-06, "loss": 0.0062, "step": 17965 }, { "epoch": 5.58, "learning_rate": 2.816963392678536e-06, "loss": 0.0023, "step": 17966 }, { "epoch": 5.58, "learning_rate": 2.8165633126625324e-06, "loss": 0.0033, "step": 17967 }, { "epoch": 5.58, "learning_rate": 2.8161632326465295e-06, "loss": 0.0031, "step": 17968 }, { "epoch": 5.58, "learning_rate": 2.815763152630526e-06, "loss": 0.0107, "step": 17969 }, { "epoch": 5.58, "learning_rate": 2.8153630726145233e-06, "loss": 0.0027, "step": 17970 }, { "epoch": 5.58, "learning_rate": 2.81496299259852e-06, "loss": 0.0069, "step": 17971 }, { "epoch": 5.58, "learning_rate": 2.814562912582517e-06, "loss": 0.0035, "step": 17972 }, { "epoch": 5.58, "learning_rate": 2.814162832566514e-06, "loss": 0.0021, "step": 17973 }, { "epoch": 5.58, "learning_rate": 2.81376275255051e-06, "loss": 0.0063, "step": 17974 }, { "epoch": 5.58, "learning_rate": 2.813362672534507e-06, "loss": 0.0178, "step": 17975 }, { "epoch": 5.58, "learning_rate": 2.812962592518504e-06, "loss": 0.0143, "step": 17976 }, { "epoch": 5.58, "learning_rate": 2.812562512502501e-06, "loss": 0.0155, "step": 17977 }, { "epoch": 5.58, "learning_rate": 2.8121624324864976e-06, "loss": 0.0125, "step": 17978 }, { "epoch": 5.58, "learning_rate": 2.8117623524704948e-06, "loss": 0.0039, "step": 17979 }, { "epoch": 5.58, "learning_rate": 2.811362272454491e-06, "loss": 0.0096, "step": 17980 }, { "epoch": 5.58, "learning_rate": 2.8109621924384877e-06, "loss": 0.0074, "step": 17981 }, { "epoch": 5.58, "learning_rate": 2.810562112422485e-06, "loss": 0.0139, "step": 17982 }, { "epoch": 5.58, "learning_rate": 2.8101620324064815e-06, "loss": 0.0037, "step": 17983 }, { "epoch": 5.58, "learning_rate": 2.8097619523904786e-06, "loss": 0.0143, "step": 17984 }, { "epoch": 5.58, "learning_rate": 2.8093618723744753e-06, "loss": 0.0049, "step": 17985 }, { "epoch": 5.58, "learning_rate": 2.8089617923584716e-06, "loss": 0.0106, "step": 17986 }, { "epoch": 5.58, "learning_rate": 2.8085617123424687e-06, "loss": 0.0021, "step": 17987 }, { "epoch": 5.58, "learning_rate": 2.8081616323264654e-06, "loss": 0.0053, "step": 17988 }, { "epoch": 5.58, "learning_rate": 2.8077615523104625e-06, "loss": 0.0057, "step": 17989 }, { "epoch": 5.58, "learning_rate": 2.807361472294459e-06, "loss": 0.0172, "step": 17990 }, { "epoch": 5.58, "learning_rate": 2.8069613922784563e-06, "loss": 0.0199, "step": 17991 }, { "epoch": 5.58, "learning_rate": 2.806561312262453e-06, "loss": 0.0019, "step": 17992 }, { "epoch": 5.58, "learning_rate": 2.806161232246449e-06, "loss": 0.0057, "step": 17993 }, { "epoch": 5.58, "learning_rate": 2.8057611522304463e-06, "loss": 0.0039, "step": 17994 }, { "epoch": 5.59, "learning_rate": 2.805361072214443e-06, "loss": 0.0108, "step": 17995 }, { "epoch": 5.59, "learning_rate": 2.80496099219844e-06, "loss": 0.007, "step": 17996 }, { "epoch": 5.59, "learning_rate": 2.804560912182437e-06, "loss": 0.0057, "step": 17997 }, { "epoch": 5.59, "learning_rate": 2.804160832166434e-06, "loss": 0.0313, "step": 17998 }, { "epoch": 5.59, "learning_rate": 2.80376075215043e-06, "loss": 0.0265, "step": 17999 }, { "epoch": 5.59, "learning_rate": 2.803360672134427e-06, "loss": 0.0032, "step": 18000 }, { "epoch": 5.59, "learning_rate": 2.802960592118424e-06, "loss": 0.0011, "step": 18001 }, { "epoch": 5.59, "learning_rate": 2.8025605121024207e-06, "loss": 0.0091, "step": 18002 }, { "epoch": 5.59, "learning_rate": 2.8021604320864178e-06, "loss": 0.0043, "step": 18003 }, { "epoch": 5.59, "learning_rate": 2.8017603520704145e-06, "loss": 0.0077, "step": 18004 }, { "epoch": 5.59, "learning_rate": 2.8013602720544107e-06, "loss": 0.0222, "step": 18005 }, { "epoch": 5.59, "learning_rate": 2.800960192038408e-06, "loss": 0.0135, "step": 18006 }, { "epoch": 5.59, "learning_rate": 2.8005601120224045e-06, "loss": 0.0027, "step": 18007 }, { "epoch": 5.59, "learning_rate": 2.8001600320064016e-06, "loss": 0.0162, "step": 18008 }, { "epoch": 5.59, "learning_rate": 2.7997599519903983e-06, "loss": 0.002, "step": 18009 }, { "epoch": 5.59, "learning_rate": 2.7993598719743954e-06, "loss": 0.1022, "step": 18010 }, { "epoch": 5.59, "learning_rate": 2.798959791958392e-06, "loss": 0.014, "step": 18011 }, { "epoch": 5.59, "learning_rate": 2.7985597119423884e-06, "loss": 0.0472, "step": 18012 }, { "epoch": 5.59, "learning_rate": 2.7981596319263855e-06, "loss": 0.0116, "step": 18013 }, { "epoch": 5.59, "learning_rate": 2.797759551910382e-06, "loss": 0.0089, "step": 18014 }, { "epoch": 5.59, "learning_rate": 2.7973594718943793e-06, "loss": 0.0073, "step": 18015 }, { "epoch": 5.59, "learning_rate": 2.796959391878376e-06, "loss": 0.0062, "step": 18016 }, { "epoch": 5.59, "learning_rate": 2.796559311862373e-06, "loss": 0.0116, "step": 18017 }, { "epoch": 5.59, "learning_rate": 2.7961592318463693e-06, "loss": 0.0053, "step": 18018 }, { "epoch": 5.59, "learning_rate": 2.795759151830366e-06, "loss": 0.0072, "step": 18019 }, { "epoch": 5.59, "learning_rate": 2.795359071814363e-06, "loss": 0.0086, "step": 18020 }, { "epoch": 5.59, "learning_rate": 2.79495899179836e-06, "loss": 0.0029, "step": 18021 }, { "epoch": 5.59, "learning_rate": 2.794558911782357e-06, "loss": 0.0066, "step": 18022 }, { "epoch": 5.59, "learning_rate": 2.7941588317663536e-06, "loss": 0.0043, "step": 18023 }, { "epoch": 5.59, "learning_rate": 2.79375875175035e-06, "loss": 0.0126, "step": 18024 }, { "epoch": 5.59, "learning_rate": 2.793358671734347e-06, "loss": 0.0055, "step": 18025 }, { "epoch": 5.59, "learning_rate": 2.7929585917183437e-06, "loss": 0.0033, "step": 18026 }, { "epoch": 5.6, "learning_rate": 2.7925585117023408e-06, "loss": 0.0107, "step": 18027 }, { "epoch": 5.6, "learning_rate": 2.7921584316863375e-06, "loss": 0.0025, "step": 18028 }, { "epoch": 5.6, "learning_rate": 2.7917583516703346e-06, "loss": 0.0103, "step": 18029 }, { "epoch": 5.6, "learning_rate": 2.791358271654331e-06, "loss": 0.0075, "step": 18030 }, { "epoch": 5.6, "learning_rate": 2.7909581916383275e-06, "loss": 0.0084, "step": 18031 }, { "epoch": 5.6, "learning_rate": 2.7905581116223246e-06, "loss": 0.0139, "step": 18032 }, { "epoch": 5.6, "learning_rate": 2.7901580316063213e-06, "loss": 0.0095, "step": 18033 }, { "epoch": 5.6, "learning_rate": 2.7897579515903184e-06, "loss": 0.0026, "step": 18034 }, { "epoch": 5.6, "learning_rate": 2.789357871574315e-06, "loss": 0.0246, "step": 18035 }, { "epoch": 5.6, "learning_rate": 2.7889577915583122e-06, "loss": 0.0119, "step": 18036 }, { "epoch": 5.6, "learning_rate": 2.7885577115423085e-06, "loss": 0.0063, "step": 18037 }, { "epoch": 5.6, "learning_rate": 2.788157631526305e-06, "loss": 0.0036, "step": 18038 }, { "epoch": 5.6, "learning_rate": 2.7877575515103023e-06, "loss": 0.017, "step": 18039 }, { "epoch": 5.6, "learning_rate": 2.787357471494299e-06, "loss": 0.0093, "step": 18040 }, { "epoch": 5.6, "learning_rate": 2.786957391478296e-06, "loss": 0.0034, "step": 18041 }, { "epoch": 5.6, "learning_rate": 2.7865573114622928e-06, "loss": 0.0078, "step": 18042 }, { "epoch": 5.6, "learning_rate": 2.7861572314462895e-06, "loss": 0.0975, "step": 18043 }, { "epoch": 5.6, "learning_rate": 2.785757151430286e-06, "loss": 0.0211, "step": 18044 }, { "epoch": 5.6, "learning_rate": 2.785357071414283e-06, "loss": 0.0163, "step": 18045 }, { "epoch": 5.6, "learning_rate": 2.78495699139828e-06, "loss": 0.005, "step": 18046 }, { "epoch": 5.6, "learning_rate": 2.7845569113822766e-06, "loss": 0.0132, "step": 18047 }, { "epoch": 5.6, "learning_rate": 2.7841568313662737e-06, "loss": 0.0085, "step": 18048 }, { "epoch": 5.6, "learning_rate": 2.78375675135027e-06, "loss": 0.0092, "step": 18049 }, { "epoch": 5.6, "learning_rate": 2.783356671334267e-06, "loss": 0.0024, "step": 18050 }, { "epoch": 5.6, "learning_rate": 2.782956591318264e-06, "loss": 0.0069, "step": 18051 }, { "epoch": 5.6, "learning_rate": 2.7825565113022605e-06, "loss": 0.0023, "step": 18052 }, { "epoch": 5.6, "learning_rate": 2.7821564312862576e-06, "loss": 0.0036, "step": 18053 }, { "epoch": 5.6, "learning_rate": 2.7817563512702543e-06, "loss": 0.0056, "step": 18054 }, { "epoch": 5.6, "learning_rate": 2.7813562712542514e-06, "loss": 0.0154, "step": 18055 }, { "epoch": 5.6, "learning_rate": 2.7809561912382476e-06, "loss": 0.0192, "step": 18056 }, { "epoch": 5.6, "learning_rate": 2.7805561112222448e-06, "loss": 0.0017, "step": 18057 }, { "epoch": 5.6, "learning_rate": 2.7801560312062414e-06, "loss": 0.0056, "step": 18058 }, { "epoch": 5.61, "learning_rate": 2.779755951190238e-06, "loss": 0.0804, "step": 18059 }, { "epoch": 5.61, "learning_rate": 2.7793558711742352e-06, "loss": 0.0161, "step": 18060 }, { "epoch": 5.61, "learning_rate": 2.778955791158232e-06, "loss": 0.0121, "step": 18061 }, { "epoch": 5.61, "learning_rate": 2.7785557111422286e-06, "loss": 0.0046, "step": 18062 }, { "epoch": 5.61, "learning_rate": 2.7781556311262253e-06, "loss": 0.0435, "step": 18063 }, { "epoch": 5.61, "learning_rate": 2.7777555511102224e-06, "loss": 0.0514, "step": 18064 }, { "epoch": 5.61, "learning_rate": 2.777355471094219e-06, "loss": 0.0113, "step": 18065 }, { "epoch": 5.61, "learning_rate": 2.7769553910782158e-06, "loss": 0.0051, "step": 18066 }, { "epoch": 5.61, "learning_rate": 2.776555311062213e-06, "loss": 0.0087, "step": 18067 }, { "epoch": 5.61, "learning_rate": 2.776155231046209e-06, "loss": 0.0023, "step": 18068 }, { "epoch": 5.61, "learning_rate": 2.7757551510302063e-06, "loss": 0.0019, "step": 18069 }, { "epoch": 5.61, "learning_rate": 2.775355071014203e-06, "loss": 0.0054, "step": 18070 }, { "epoch": 5.61, "learning_rate": 2.7749549909982e-06, "loss": 0.0059, "step": 18071 }, { "epoch": 5.61, "learning_rate": 2.7745549109821967e-06, "loss": 0.0046, "step": 18072 }, { "epoch": 5.61, "learning_rate": 2.774154830966194e-06, "loss": 0.0087, "step": 18073 }, { "epoch": 5.61, "learning_rate": 2.7737547509501905e-06, "loss": 0.0392, "step": 18074 }, { "epoch": 5.61, "learning_rate": 2.773354670934187e-06, "loss": 0.0082, "step": 18075 }, { "epoch": 5.61, "learning_rate": 2.772954590918184e-06, "loss": 0.011, "step": 18076 }, { "epoch": 5.61, "learning_rate": 2.7725545109021806e-06, "loss": 0.005, "step": 18077 }, { "epoch": 5.61, "learning_rate": 2.7721544308861777e-06, "loss": 0.005, "step": 18078 }, { "epoch": 5.61, "learning_rate": 2.7717543508701744e-06, "loss": 0.0544, "step": 18079 }, { "epoch": 5.61, "learning_rate": 2.7713542708541715e-06, "loss": 0.003, "step": 18080 }, { "epoch": 5.61, "learning_rate": 2.7709541908381678e-06, "loss": 0.0218, "step": 18081 }, { "epoch": 5.61, "learning_rate": 2.7705541108221645e-06, "loss": 0.0047, "step": 18082 }, { "epoch": 5.61, "learning_rate": 2.7701540308061616e-06, "loss": 0.0023, "step": 18083 }, { "epoch": 5.61, "learning_rate": 2.7697539507901582e-06, "loss": 0.0035, "step": 18084 }, { "epoch": 5.61, "learning_rate": 2.7693538707741554e-06, "loss": 0.0087, "step": 18085 }, { "epoch": 5.61, "learning_rate": 2.768953790758152e-06, "loss": 0.0035, "step": 18086 }, { "epoch": 5.61, "learning_rate": 2.7685537107421483e-06, "loss": 0.0048, "step": 18087 }, { "epoch": 5.61, "learning_rate": 2.7681536307261454e-06, "loss": 0.0032, "step": 18088 }, { "epoch": 5.61, "learning_rate": 2.767753550710142e-06, "loss": 0.0059, "step": 18089 }, { "epoch": 5.61, "learning_rate": 2.7673534706941392e-06, "loss": 0.0054, "step": 18090 }, { "epoch": 5.62, "learning_rate": 2.766953390678136e-06, "loss": 0.0119, "step": 18091 }, { "epoch": 5.62, "learning_rate": 2.766553310662133e-06, "loss": 0.0033, "step": 18092 }, { "epoch": 5.62, "learning_rate": 2.7661532306461297e-06, "loss": 0.0071, "step": 18093 }, { "epoch": 5.62, "learning_rate": 2.765753150630126e-06, "loss": 0.0098, "step": 18094 }, { "epoch": 5.62, "learning_rate": 2.765353070614123e-06, "loss": 0.0286, "step": 18095 }, { "epoch": 5.62, "learning_rate": 2.7649529905981198e-06, "loss": 0.0065, "step": 18096 }, { "epoch": 5.62, "learning_rate": 2.764552910582117e-06, "loss": 0.008, "step": 18097 }, { "epoch": 5.62, "learning_rate": 2.7641528305661136e-06, "loss": 0.0046, "step": 18098 }, { "epoch": 5.62, "learning_rate": 2.7637527505501107e-06, "loss": 0.0376, "step": 18099 }, { "epoch": 5.62, "learning_rate": 2.763352670534107e-06, "loss": 0.0039, "step": 18100 }, { "epoch": 5.62, "learning_rate": 2.7629525905181036e-06, "loss": 0.0038, "step": 18101 }, { "epoch": 5.62, "learning_rate": 2.7625525105021007e-06, "loss": 0.0545, "step": 18102 }, { "epoch": 5.62, "learning_rate": 2.7621524304860974e-06, "loss": 0.0036, "step": 18103 }, { "epoch": 5.62, "learning_rate": 2.7617523504700945e-06, "loss": 0.009, "step": 18104 }, { "epoch": 5.62, "learning_rate": 2.761352270454091e-06, "loss": 0.0115, "step": 18105 }, { "epoch": 5.62, "learning_rate": 2.7609521904380875e-06, "loss": 0.0114, "step": 18106 }, { "epoch": 5.62, "learning_rate": 2.7605521104220846e-06, "loss": 0.0063, "step": 18107 }, { "epoch": 5.62, "learning_rate": 2.7601520304060813e-06, "loss": 0.0044, "step": 18108 }, { "epoch": 5.62, "learning_rate": 2.7597519503900784e-06, "loss": 0.0109, "step": 18109 }, { "epoch": 5.62, "learning_rate": 2.759351870374075e-06, "loss": 0.0265, "step": 18110 }, { "epoch": 5.62, "learning_rate": 2.758951790358072e-06, "loss": 0.0036, "step": 18111 }, { "epoch": 5.62, "learning_rate": 2.7585517103420684e-06, "loss": 0.0117, "step": 18112 }, { "epoch": 5.62, "learning_rate": 2.758151630326065e-06, "loss": 0.011, "step": 18113 }, { "epoch": 5.62, "learning_rate": 2.7577515503100622e-06, "loss": 0.0106, "step": 18114 }, { "epoch": 5.62, "learning_rate": 2.757351470294059e-06, "loss": 0.0081, "step": 18115 }, { "epoch": 5.62, "learning_rate": 2.756951390278056e-06, "loss": 0.0096, "step": 18116 }, { "epoch": 5.62, "learning_rate": 2.7565513102620527e-06, "loss": 0.0155, "step": 18117 }, { "epoch": 5.62, "learning_rate": 2.75615123024605e-06, "loss": 0.0544, "step": 18118 }, { "epoch": 5.62, "learning_rate": 2.755751150230046e-06, "loss": 0.0104, "step": 18119 }, { "epoch": 5.62, "learning_rate": 2.7553510702140428e-06, "loss": 0.0028, "step": 18120 }, { "epoch": 5.62, "learning_rate": 2.75495099019804e-06, "loss": 0.004, "step": 18121 }, { "epoch": 5.62, "learning_rate": 2.7545509101820366e-06, "loss": 0.0782, "step": 18122 }, { "epoch": 5.62, "learning_rate": 2.7541508301660337e-06, "loss": 0.0927, "step": 18123 }, { "epoch": 5.63, "learning_rate": 2.7537507501500304e-06, "loss": 0.004, "step": 18124 }, { "epoch": 5.63, "learning_rate": 2.7533506701340266e-06, "loss": 0.0075, "step": 18125 }, { "epoch": 5.63, "learning_rate": 2.7529505901180237e-06, "loss": 0.0077, "step": 18126 }, { "epoch": 5.63, "learning_rate": 2.7525505101020204e-06, "loss": 0.0281, "step": 18127 }, { "epoch": 5.63, "learning_rate": 2.7521504300860175e-06, "loss": 0.0043, "step": 18128 }, { "epoch": 5.63, "learning_rate": 2.7517503500700142e-06, "loss": 0.0044, "step": 18129 }, { "epoch": 5.63, "learning_rate": 2.7513502700540113e-06, "loss": 0.0067, "step": 18130 }, { "epoch": 5.63, "learning_rate": 2.7509501900380076e-06, "loss": 0.0091, "step": 18131 }, { "epoch": 5.63, "learning_rate": 2.7505501100220043e-06, "loss": 0.0021, "step": 18132 }, { "epoch": 5.63, "learning_rate": 2.7501500300060014e-06, "loss": 0.0069, "step": 18133 }, { "epoch": 5.63, "learning_rate": 2.749749949989998e-06, "loss": 0.0088, "step": 18134 }, { "epoch": 5.63, "learning_rate": 2.749349869973995e-06, "loss": 0.0082, "step": 18135 }, { "epoch": 5.63, "learning_rate": 2.748949789957992e-06, "loss": 0.0061, "step": 18136 }, { "epoch": 5.63, "learning_rate": 2.748549709941989e-06, "loss": 0.0417, "step": 18137 }, { "epoch": 5.63, "learning_rate": 2.7481496299259852e-06, "loss": 0.0033, "step": 18138 }, { "epoch": 5.63, "learning_rate": 2.747749549909982e-06, "loss": 0.006, "step": 18139 }, { "epoch": 5.63, "learning_rate": 2.747349469893979e-06, "loss": 0.002, "step": 18140 }, { "epoch": 5.63, "learning_rate": 2.7469493898779757e-06, "loss": 0.0178, "step": 18141 }, { "epoch": 5.63, "learning_rate": 2.746549309861973e-06, "loss": 0.0152, "step": 18142 }, { "epoch": 5.63, "learning_rate": 2.7461492298459695e-06, "loss": 0.004, "step": 18143 }, { "epoch": 5.63, "learning_rate": 2.7457491498299658e-06, "loss": 0.0037, "step": 18144 }, { "epoch": 5.63, "learning_rate": 2.745349069813963e-06, "loss": 0.0019, "step": 18145 }, { "epoch": 5.63, "learning_rate": 2.7449489897979596e-06, "loss": 0.0507, "step": 18146 }, { "epoch": 5.63, "learning_rate": 2.7445489097819567e-06, "loss": 0.0146, "step": 18147 }, { "epoch": 5.63, "learning_rate": 2.7441488297659534e-06, "loss": 0.0031, "step": 18148 }, { "epoch": 5.63, "learning_rate": 2.7437487497499505e-06, "loss": 0.003, "step": 18149 }, { "epoch": 5.63, "learning_rate": 2.7433486697339467e-06, "loss": 0.0025, "step": 18150 }, { "epoch": 5.63, "learning_rate": 2.742948589717944e-06, "loss": 0.0743, "step": 18151 }, { "epoch": 5.63, "learning_rate": 2.7425485097019405e-06, "loss": 0.0042, "step": 18152 }, { "epoch": 5.63, "learning_rate": 2.7421484296859372e-06, "loss": 0.0056, "step": 18153 }, { "epoch": 5.63, "learning_rate": 2.7417483496699343e-06, "loss": 0.0034, "step": 18154 }, { "epoch": 5.63, "learning_rate": 2.741348269653931e-06, "loss": 0.0218, "step": 18155 }, { "epoch": 5.64, "learning_rate": 2.740948189637928e-06, "loss": 0.0055, "step": 18156 }, { "epoch": 5.64, "learning_rate": 2.7405481096219244e-06, "loss": 0.0224, "step": 18157 }, { "epoch": 5.64, "learning_rate": 2.7401480296059215e-06, "loss": 0.0024, "step": 18158 }, { "epoch": 5.64, "learning_rate": 2.739747949589918e-06, "loss": 0.0033, "step": 18159 }, { "epoch": 5.64, "learning_rate": 2.739347869573915e-06, "loss": 0.0083, "step": 18160 }, { "epoch": 5.64, "learning_rate": 2.738947789557912e-06, "loss": 0.0037, "step": 18161 }, { "epoch": 5.64, "learning_rate": 2.7385477095419087e-06, "loss": 0.0098, "step": 18162 }, { "epoch": 5.64, "learning_rate": 2.7381476295259054e-06, "loss": 0.0085, "step": 18163 }, { "epoch": 5.64, "learning_rate": 2.737747549509902e-06, "loss": 0.0022, "step": 18164 }, { "epoch": 5.64, "learning_rate": 2.737347469493899e-06, "loss": 0.0025, "step": 18165 }, { "epoch": 5.64, "learning_rate": 2.736947389477896e-06, "loss": 0.0026, "step": 18166 }, { "epoch": 5.64, "learning_rate": 2.7365473094618925e-06, "loss": 0.0151, "step": 18167 }, { "epoch": 5.64, "learning_rate": 2.7361472294458896e-06, "loss": 0.0018, "step": 18168 }, { "epoch": 5.64, "learning_rate": 2.735747149429886e-06, "loss": 0.0031, "step": 18169 }, { "epoch": 5.64, "learning_rate": 2.735347069413883e-06, "loss": 0.0037, "step": 18170 }, { "epoch": 5.64, "learning_rate": 2.7349469893978797e-06, "loss": 0.0186, "step": 18171 }, { "epoch": 5.64, "learning_rate": 2.734546909381877e-06, "loss": 0.0031, "step": 18172 }, { "epoch": 5.64, "learning_rate": 2.7341468293658735e-06, "loss": 0.0078, "step": 18173 }, { "epoch": 5.64, "learning_rate": 2.73374674934987e-06, "loss": 0.0093, "step": 18174 }, { "epoch": 5.64, "learning_rate": 2.733346669333867e-06, "loss": 0.0563, "step": 18175 }, { "epoch": 5.64, "learning_rate": 2.7329465893178636e-06, "loss": 0.0077, "step": 18176 }, { "epoch": 5.64, "learning_rate": 2.7325465093018607e-06, "loss": 0.0311, "step": 18177 }, { "epoch": 5.64, "learning_rate": 2.7321464292858573e-06, "loss": 0.0071, "step": 18178 }, { "epoch": 5.64, "learning_rate": 2.7317463492698545e-06, "loss": 0.0097, "step": 18179 }, { "epoch": 5.64, "learning_rate": 2.731346269253851e-06, "loss": 0.0101, "step": 18180 }, { "epoch": 5.64, "learning_rate": 2.7309461892378483e-06, "loss": 0.0031, "step": 18181 }, { "epoch": 5.64, "learning_rate": 2.7305461092218445e-06, "loss": 0.0117, "step": 18182 }, { "epoch": 5.64, "learning_rate": 2.730146029205841e-06, "loss": 0.0095, "step": 18183 }, { "epoch": 5.64, "learning_rate": 2.7297459491898383e-06, "loss": 0.007, "step": 18184 }, { "epoch": 5.64, "learning_rate": 2.729345869173835e-06, "loss": 0.003, "step": 18185 }, { "epoch": 5.64, "learning_rate": 2.728945789157832e-06, "loss": 0.0051, "step": 18186 }, { "epoch": 5.64, "learning_rate": 2.728545709141829e-06, "loss": 0.0196, "step": 18187 }, { "epoch": 5.65, "learning_rate": 2.728145629125825e-06, "loss": 0.0067, "step": 18188 }, { "epoch": 5.65, "learning_rate": 2.727745549109822e-06, "loss": 0.0541, "step": 18189 }, { "epoch": 5.65, "learning_rate": 2.727345469093819e-06, "loss": 0.0044, "step": 18190 }, { "epoch": 5.65, "learning_rate": 2.726945389077816e-06, "loss": 0.0046, "step": 18191 }, { "epoch": 5.65, "learning_rate": 2.7265453090618126e-06, "loss": 0.0055, "step": 18192 }, { "epoch": 5.65, "learning_rate": 2.7261452290458098e-06, "loss": 0.0062, "step": 18193 }, { "epoch": 5.65, "learning_rate": 2.725745149029806e-06, "loss": 0.0104, "step": 18194 }, { "epoch": 5.65, "learning_rate": 2.7253450690138027e-06, "loss": 0.0051, "step": 18195 }, { "epoch": 5.65, "learning_rate": 2.7249449889978e-06, "loss": 0.0165, "step": 18196 }, { "epoch": 5.65, "learning_rate": 2.7245449089817965e-06, "loss": 0.0082, "step": 18197 }, { "epoch": 5.65, "learning_rate": 2.7241448289657936e-06, "loss": 0.0624, "step": 18198 }, { "epoch": 5.65, "learning_rate": 2.7237447489497903e-06, "loss": 0.0059, "step": 18199 }, { "epoch": 5.65, "learning_rate": 2.7233446689337874e-06, "loss": 0.0215, "step": 18200 }, { "epoch": 5.65, "learning_rate": 2.7229445889177837e-06, "loss": 0.0715, "step": 18201 }, { "epoch": 5.65, "learning_rate": 2.7225445089017804e-06, "loss": 0.0037, "step": 18202 }, { "epoch": 5.65, "learning_rate": 2.7221444288857775e-06, "loss": 0.0048, "step": 18203 }, { "epoch": 5.65, "learning_rate": 2.721744348869774e-06, "loss": 0.0129, "step": 18204 }, { "epoch": 5.65, "learning_rate": 2.7213442688537713e-06, "loss": 0.0025, "step": 18205 }, { "epoch": 5.65, "learning_rate": 2.720944188837768e-06, "loss": 0.0089, "step": 18206 }, { "epoch": 5.65, "learning_rate": 2.7205441088217642e-06, "loss": 0.0019, "step": 18207 }, { "epoch": 5.65, "learning_rate": 2.7201440288057613e-06, "loss": 0.0474, "step": 18208 }, { "epoch": 5.65, "learning_rate": 2.719743948789758e-06, "loss": 0.0015, "step": 18209 }, { "epoch": 5.65, "learning_rate": 2.719343868773755e-06, "loss": 0.0044, "step": 18210 }, { "epoch": 5.65, "learning_rate": 2.718943788757752e-06, "loss": 0.0071, "step": 18211 }, { "epoch": 5.65, "learning_rate": 2.718543708741749e-06, "loss": 0.0047, "step": 18212 }, { "epoch": 5.65, "learning_rate": 2.718143628725745e-06, "loss": 0.0041, "step": 18213 }, { "epoch": 5.65, "learning_rate": 2.717743548709742e-06, "loss": 0.0106, "step": 18214 }, { "epoch": 5.65, "learning_rate": 2.717343468693739e-06, "loss": 0.0069, "step": 18215 }, { "epoch": 5.65, "learning_rate": 2.7169433886777357e-06, "loss": 0.0029, "step": 18216 }, { "epoch": 5.65, "learning_rate": 2.7165433086617328e-06, "loss": 0.0038, "step": 18217 }, { "epoch": 5.65, "learning_rate": 2.7161432286457295e-06, "loss": 0.0516, "step": 18218 }, { "epoch": 5.65, "learning_rate": 2.7157431486297266e-06, "loss": 0.0044, "step": 18219 }, { "epoch": 5.66, "learning_rate": 2.715343068613723e-06, "loss": 0.013, "step": 18220 }, { "epoch": 5.66, "learning_rate": 2.7149429885977195e-06, "loss": 0.0601, "step": 18221 }, { "epoch": 5.66, "learning_rate": 2.7145429085817166e-06, "loss": 0.0088, "step": 18222 }, { "epoch": 5.66, "learning_rate": 2.7141428285657133e-06, "loss": 0.0026, "step": 18223 }, { "epoch": 5.66, "learning_rate": 2.7137427485497104e-06, "loss": 0.0042, "step": 18224 }, { "epoch": 5.66, "learning_rate": 2.713342668533707e-06, "loss": 0.0071, "step": 18225 }, { "epoch": 5.66, "learning_rate": 2.7129425885177034e-06, "loss": 0.017, "step": 18226 }, { "epoch": 5.66, "learning_rate": 2.7125425085017005e-06, "loss": 0.0034, "step": 18227 }, { "epoch": 5.66, "learning_rate": 2.712142428485697e-06, "loss": 0.0052, "step": 18228 }, { "epoch": 5.66, "learning_rate": 2.7117423484696943e-06, "loss": 0.014, "step": 18229 }, { "epoch": 5.66, "learning_rate": 2.711342268453691e-06, "loss": 0.0049, "step": 18230 }, { "epoch": 5.66, "learning_rate": 2.710942188437688e-06, "loss": 0.0032, "step": 18231 }, { "epoch": 5.66, "learning_rate": 2.7105421084216843e-06, "loss": 0.0052, "step": 18232 }, { "epoch": 5.66, "learning_rate": 2.710142028405681e-06, "loss": 0.0066, "step": 18233 }, { "epoch": 5.66, "learning_rate": 2.709741948389678e-06, "loss": 0.0018, "step": 18234 }, { "epoch": 5.66, "learning_rate": 2.709341868373675e-06, "loss": 0.0024, "step": 18235 }, { "epoch": 5.66, "learning_rate": 2.708941788357672e-06, "loss": 0.0085, "step": 18236 }, { "epoch": 5.66, "learning_rate": 2.7085417083416686e-06, "loss": 0.0063, "step": 18237 }, { "epoch": 5.66, "learning_rate": 2.7081416283256657e-06, "loss": 0.02, "step": 18238 }, { "epoch": 5.66, "learning_rate": 2.707741548309662e-06, "loss": 0.0042, "step": 18239 }, { "epoch": 5.66, "learning_rate": 2.7073414682936587e-06, "loss": 0.0051, "step": 18240 }, { "epoch": 5.66, "learning_rate": 2.7069413882776558e-06, "loss": 0.0122, "step": 18241 }, { "epoch": 5.66, "learning_rate": 2.7065413082616525e-06, "loss": 0.0106, "step": 18242 }, { "epoch": 5.66, "learning_rate": 2.7061412282456496e-06, "loss": 0.0067, "step": 18243 }, { "epoch": 5.66, "learning_rate": 2.7057411482296463e-06, "loss": 0.0045, "step": 18244 }, { "epoch": 5.66, "learning_rate": 2.7053410682136425e-06, "loss": 0.0085, "step": 18245 }, { "epoch": 5.66, "learning_rate": 2.7049409881976396e-06, "loss": 0.0106, "step": 18246 }, { "epoch": 5.66, "learning_rate": 2.7045409081816363e-06, "loss": 0.0025, "step": 18247 }, { "epoch": 5.66, "learning_rate": 2.7041408281656334e-06, "loss": 0.011, "step": 18248 }, { "epoch": 5.66, "learning_rate": 2.70374074814963e-06, "loss": 0.057, "step": 18249 }, { "epoch": 5.66, "learning_rate": 2.7033406681336272e-06, "loss": 0.0024, "step": 18250 }, { "epoch": 5.66, "learning_rate": 2.7029405881176235e-06, "loss": 0.0062, "step": 18251 }, { "epoch": 5.67, "learning_rate": 2.70254050810162e-06, "loss": 0.0054, "step": 18252 }, { "epoch": 5.67, "learning_rate": 2.7021404280856173e-06, "loss": 0.0016, "step": 18253 }, { "epoch": 5.67, "learning_rate": 2.701740348069614e-06, "loss": 0.0053, "step": 18254 }, { "epoch": 5.67, "learning_rate": 2.701340268053611e-06, "loss": 0.0045, "step": 18255 }, { "epoch": 5.67, "learning_rate": 2.7009401880376078e-06, "loss": 0.0139, "step": 18256 }, { "epoch": 5.67, "learning_rate": 2.7005401080216045e-06, "loss": 0.0027, "step": 18257 }, { "epoch": 5.67, "learning_rate": 2.700140028005601e-06, "loss": 0.0082, "step": 18258 }, { "epoch": 5.67, "learning_rate": 2.6997399479895983e-06, "loss": 0.02, "step": 18259 }, { "epoch": 5.67, "learning_rate": 2.699339867973595e-06, "loss": 0.0084, "step": 18260 }, { "epoch": 5.67, "learning_rate": 2.6989397879575916e-06, "loss": 0.0154, "step": 18261 }, { "epoch": 5.67, "learning_rate": 2.6985397079415887e-06, "loss": 0.0031, "step": 18262 }, { "epoch": 5.67, "learning_rate": 2.6981396279255854e-06, "loss": 0.0015, "step": 18263 }, { "epoch": 5.67, "learning_rate": 2.697739547909582e-06, "loss": 0.0068, "step": 18264 }, { "epoch": 5.67, "learning_rate": 2.697339467893579e-06, "loss": 0.0118, "step": 18265 }, { "epoch": 5.67, "learning_rate": 2.696939387877576e-06, "loss": 0.0047, "step": 18266 }, { "epoch": 5.67, "learning_rate": 2.6965393078615726e-06, "loss": 0.0054, "step": 18267 }, { "epoch": 5.67, "learning_rate": 2.6961392278455693e-06, "loss": 0.0018, "step": 18268 }, { "epoch": 5.67, "learning_rate": 2.6957391478295664e-06, "loss": 0.0339, "step": 18269 }, { "epoch": 5.67, "learning_rate": 2.6953390678135626e-06, "loss": 0.0041, "step": 18270 }, { "epoch": 5.67, "learning_rate": 2.6949389877975598e-06, "loss": 0.0018, "step": 18271 }, { "epoch": 5.67, "learning_rate": 2.6945389077815564e-06, "loss": 0.0028, "step": 18272 }, { "epoch": 5.67, "learning_rate": 2.6941388277655536e-06, "loss": 0.0111, "step": 18273 }, { "epoch": 5.67, "learning_rate": 2.6937387477495502e-06, "loss": 0.018, "step": 18274 }, { "epoch": 5.67, "learning_rate": 2.693338667733547e-06, "loss": 0.0057, "step": 18275 }, { "epoch": 5.67, "learning_rate": 2.6929385877175436e-06, "loss": 0.0054, "step": 18276 }, { "epoch": 5.67, "learning_rate": 2.6925385077015403e-06, "loss": 0.0071, "step": 18277 }, { "epoch": 5.67, "learning_rate": 2.6921384276855374e-06, "loss": 0.0025, "step": 18278 }, { "epoch": 5.67, "learning_rate": 2.691738347669534e-06, "loss": 0.0072, "step": 18279 }, { "epoch": 5.67, "learning_rate": 2.691338267653531e-06, "loss": 0.0013, "step": 18280 }, { "epoch": 5.67, "learning_rate": 2.690938187637528e-06, "loss": 0.0023, "step": 18281 }, { "epoch": 5.67, "learning_rate": 2.6905381076215246e-06, "loss": 0.0958, "step": 18282 }, { "epoch": 5.67, "learning_rate": 2.6901380276055213e-06, "loss": 0.0665, "step": 18283 }, { "epoch": 5.67, "learning_rate": 2.689737947589518e-06, "loss": 0.0223, "step": 18284 }, { "epoch": 5.68, "learning_rate": 2.689337867573515e-06, "loss": 0.0034, "step": 18285 }, { "epoch": 5.68, "learning_rate": 2.6889377875575117e-06, "loss": 0.0148, "step": 18286 }, { "epoch": 5.68, "learning_rate": 2.688537707541509e-06, "loss": 0.0225, "step": 18287 }, { "epoch": 5.68, "learning_rate": 2.6881376275255055e-06, "loss": 0.0031, "step": 18288 }, { "epoch": 5.68, "learning_rate": 2.687737547509502e-06, "loss": 0.0043, "step": 18289 }, { "epoch": 5.68, "learning_rate": 2.687337467493499e-06, "loss": 0.0073, "step": 18290 }, { "epoch": 5.68, "learning_rate": 2.6869373874774956e-06, "loss": 0.0145, "step": 18291 }, { "epoch": 5.68, "learning_rate": 2.6865373074614927e-06, "loss": 0.007, "step": 18292 }, { "epoch": 5.68, "learning_rate": 2.6861372274454894e-06, "loss": 0.0011, "step": 18293 }, { "epoch": 5.68, "learning_rate": 2.6857371474294865e-06, "loss": 0.0222, "step": 18294 }, { "epoch": 5.68, "learning_rate": 2.6853370674134828e-06, "loss": 0.0076, "step": 18295 }, { "epoch": 5.68, "learning_rate": 2.6849369873974795e-06, "loss": 0.0037, "step": 18296 }, { "epoch": 5.68, "learning_rate": 2.6845369073814766e-06, "loss": 0.0089, "step": 18297 }, { "epoch": 5.68, "learning_rate": 2.6841368273654733e-06, "loss": 0.0043, "step": 18298 }, { "epoch": 5.68, "learning_rate": 2.6837367473494704e-06, "loss": 0.0085, "step": 18299 }, { "epoch": 5.68, "learning_rate": 2.683336667333467e-06, "loss": 0.0042, "step": 18300 }, { "epoch": 5.68, "learning_rate": 2.682936587317464e-06, "loss": 0.0035, "step": 18301 }, { "epoch": 5.68, "learning_rate": 2.6825365073014604e-06, "loss": 0.004, "step": 18302 }, { "epoch": 5.68, "learning_rate": 2.682136427285457e-06, "loss": 0.0034, "step": 18303 }, { "epoch": 5.68, "learning_rate": 2.6817363472694542e-06, "loss": 0.0035, "step": 18304 }, { "epoch": 5.68, "learning_rate": 2.681336267253451e-06, "loss": 0.0044, "step": 18305 }, { "epoch": 5.68, "learning_rate": 2.680936187237448e-06, "loss": 0.0115, "step": 18306 }, { "epoch": 5.68, "learning_rate": 2.6805361072214447e-06, "loss": 0.0089, "step": 18307 }, { "epoch": 5.68, "learning_rate": 2.680136027205441e-06, "loss": 0.0116, "step": 18308 }, { "epoch": 5.68, "learning_rate": 2.679735947189438e-06, "loss": 0.0046, "step": 18309 }, { "epoch": 5.68, "learning_rate": 2.6793358671734348e-06, "loss": 0.0226, "step": 18310 }, { "epoch": 5.68, "learning_rate": 2.678935787157432e-06, "loss": 0.0065, "step": 18311 }, { "epoch": 5.68, "learning_rate": 2.6785357071414286e-06, "loss": 0.0062, "step": 18312 }, { "epoch": 5.68, "learning_rate": 2.6781356271254257e-06, "loss": 0.0031, "step": 18313 }, { "epoch": 5.68, "learning_rate": 2.677735547109422e-06, "loss": 0.0038, "step": 18314 }, { "epoch": 5.68, "learning_rate": 2.6773354670934186e-06, "loss": 0.0037, "step": 18315 }, { "epoch": 5.68, "learning_rate": 2.6769353870774157e-06, "loss": 0.0034, "step": 18316 }, { "epoch": 5.69, "learning_rate": 2.6765353070614124e-06, "loss": 0.0036, "step": 18317 }, { "epoch": 5.69, "learning_rate": 2.6761352270454095e-06, "loss": 0.0055, "step": 18318 }, { "epoch": 5.69, "learning_rate": 2.675735147029406e-06, "loss": 0.0037, "step": 18319 }, { "epoch": 5.69, "learning_rate": 2.6753350670134033e-06, "loss": 0.0258, "step": 18320 }, { "epoch": 5.69, "learning_rate": 2.6749349869973996e-06, "loss": 0.0142, "step": 18321 }, { "epoch": 5.69, "learning_rate": 2.6745349069813963e-06, "loss": 0.0122, "step": 18322 }, { "epoch": 5.69, "learning_rate": 2.6741348269653934e-06, "loss": 0.0065, "step": 18323 }, { "epoch": 5.69, "learning_rate": 2.67373474694939e-06, "loss": 0.0024, "step": 18324 }, { "epoch": 5.69, "learning_rate": 2.673334666933387e-06, "loss": 0.0039, "step": 18325 }, { "epoch": 5.69, "learning_rate": 2.672934586917384e-06, "loss": 0.0097, "step": 18326 }, { "epoch": 5.69, "learning_rate": 2.67253450690138e-06, "loss": 0.0048, "step": 18327 }, { "epoch": 5.69, "learning_rate": 2.6721344268853772e-06, "loss": 0.0036, "step": 18328 }, { "epoch": 5.69, "learning_rate": 2.671734346869374e-06, "loss": 0.006, "step": 18329 }, { "epoch": 5.69, "learning_rate": 2.671334266853371e-06, "loss": 0.008, "step": 18330 }, { "epoch": 5.69, "learning_rate": 2.6709341868373677e-06, "loss": 0.0172, "step": 18331 }, { "epoch": 5.69, "learning_rate": 2.670534106821365e-06, "loss": 0.0034, "step": 18332 }, { "epoch": 5.69, "learning_rate": 2.670134026805361e-06, "loss": 0.0046, "step": 18333 }, { "epoch": 5.69, "learning_rate": 2.6697339467893578e-06, "loss": 0.0158, "step": 18334 }, { "epoch": 5.69, "learning_rate": 2.669333866773355e-06, "loss": 0.0081, "step": 18335 }, { "epoch": 5.69, "learning_rate": 2.6689337867573516e-06, "loss": 0.0274, "step": 18336 }, { "epoch": 5.69, "learning_rate": 2.6685337067413487e-06, "loss": 0.0017, "step": 18337 }, { "epoch": 5.69, "learning_rate": 2.6681336267253454e-06, "loss": 0.0061, "step": 18338 }, { "epoch": 5.69, "learning_rate": 2.6677335467093416e-06, "loss": 0.0076, "step": 18339 }, { "epoch": 5.69, "learning_rate": 2.6673334666933387e-06, "loss": 0.0045, "step": 18340 }, { "epoch": 5.69, "learning_rate": 2.6669333866773354e-06, "loss": 0.0095, "step": 18341 }, { "epoch": 5.69, "learning_rate": 2.6665333066613325e-06, "loss": 0.0019, "step": 18342 }, { "epoch": 5.69, "learning_rate": 2.6661332266453292e-06, "loss": 0.0115, "step": 18343 }, { "epoch": 5.69, "learning_rate": 2.6657331466293263e-06, "loss": 0.0457, "step": 18344 }, { "epoch": 5.69, "learning_rate": 2.665333066613323e-06, "loss": 0.0042, "step": 18345 }, { "epoch": 5.69, "learning_rate": 2.6649329865973193e-06, "loss": 0.0236, "step": 18346 }, { "epoch": 5.69, "learning_rate": 2.6645329065813164e-06, "loss": 0.0101, "step": 18347 }, { "epoch": 5.69, "learning_rate": 2.664132826565313e-06, "loss": 0.0467, "step": 18348 }, { "epoch": 5.7, "learning_rate": 2.66373274654931e-06, "loss": 0.0118, "step": 18349 }, { "epoch": 5.7, "learning_rate": 2.663332666533307e-06, "loss": 0.0036, "step": 18350 }, { "epoch": 5.7, "learning_rate": 2.662932586517304e-06, "loss": 0.0087, "step": 18351 }, { "epoch": 5.7, "learning_rate": 2.6625325065013002e-06, "loss": 0.0058, "step": 18352 }, { "epoch": 5.7, "learning_rate": 2.662132426485297e-06, "loss": 0.0062, "step": 18353 }, { "epoch": 5.7, "learning_rate": 2.661732346469294e-06, "loss": 0.0025, "step": 18354 }, { "epoch": 5.7, "learning_rate": 2.6613322664532907e-06, "loss": 0.0051, "step": 18355 }, { "epoch": 5.7, "learning_rate": 2.660932186437288e-06, "loss": 0.0051, "step": 18356 }, { "epoch": 5.7, "learning_rate": 2.6605321064212845e-06, "loss": 0.0127, "step": 18357 }, { "epoch": 5.7, "learning_rate": 2.660132026405281e-06, "loss": 0.0042, "step": 18358 }, { "epoch": 5.7, "learning_rate": 2.659731946389278e-06, "loss": 0.007, "step": 18359 }, { "epoch": 5.7, "learning_rate": 2.6593318663732746e-06, "loss": 0.0056, "step": 18360 }, { "epoch": 5.7, "learning_rate": 2.6589317863572717e-06, "loss": 0.0019, "step": 18361 }, { "epoch": 5.7, "learning_rate": 2.6585317063412684e-06, "loss": 0.0107, "step": 18362 }, { "epoch": 5.7, "learning_rate": 2.6581316263252655e-06, "loss": 0.0139, "step": 18363 }, { "epoch": 5.7, "learning_rate": 2.657731546309262e-06, "loss": 0.0085, "step": 18364 }, { "epoch": 5.7, "learning_rate": 2.657331466293259e-06, "loss": 0.0079, "step": 18365 }, { "epoch": 5.7, "learning_rate": 2.6569313862772555e-06, "loss": 0.0132, "step": 18366 }, { "epoch": 5.7, "learning_rate": 2.6565313062612522e-06, "loss": 0.0031, "step": 18367 }, { "epoch": 5.7, "learning_rate": 2.6561312262452493e-06, "loss": 0.0038, "step": 18368 }, { "epoch": 5.7, "learning_rate": 2.655731146229246e-06, "loss": 0.0216, "step": 18369 }, { "epoch": 5.7, "learning_rate": 2.655331066213243e-06, "loss": 0.0101, "step": 18370 }, { "epoch": 5.7, "learning_rate": 2.6549309861972394e-06, "loss": 0.0063, "step": 18371 }, { "epoch": 5.7, "learning_rate": 2.6545309061812365e-06, "loss": 0.0013, "step": 18372 }, { "epoch": 5.7, "learning_rate": 2.654130826165233e-06, "loss": 0.021, "step": 18373 }, { "epoch": 5.7, "learning_rate": 2.6537307461492303e-06, "loss": 0.0017, "step": 18374 }, { "epoch": 5.7, "learning_rate": 2.653330666133227e-06, "loss": 0.0054, "step": 18375 }, { "epoch": 5.7, "learning_rate": 2.6529305861172237e-06, "loss": 0.0189, "step": 18376 }, { "epoch": 5.7, "learning_rate": 2.6525305061012204e-06, "loss": 0.027, "step": 18377 }, { "epoch": 5.7, "learning_rate": 2.652130426085217e-06, "loss": 0.0082, "step": 18378 }, { "epoch": 5.7, "learning_rate": 2.651730346069214e-06, "loss": 0.0045, "step": 18379 }, { "epoch": 5.7, "learning_rate": 2.651330266053211e-06, "loss": 0.0406, "step": 18380 }, { "epoch": 5.71, "learning_rate": 2.650930186037208e-06, "loss": 0.0105, "step": 18381 }, { "epoch": 5.71, "learning_rate": 2.6505301060212046e-06, "loss": 0.0155, "step": 18382 }, { "epoch": 5.71, "learning_rate": 2.6501300260052013e-06, "loss": 0.0057, "step": 18383 }, { "epoch": 5.71, "learning_rate": 2.649729945989198e-06, "loss": 0.0059, "step": 18384 }, { "epoch": 5.71, "learning_rate": 2.6493298659731947e-06, "loss": 0.0036, "step": 18385 }, { "epoch": 5.71, "learning_rate": 2.648929785957192e-06, "loss": 0.0072, "step": 18386 }, { "epoch": 5.71, "learning_rate": 2.6485297059411885e-06, "loss": 0.0127, "step": 18387 }, { "epoch": 5.71, "learning_rate": 2.6481296259251856e-06, "loss": 0.0053, "step": 18388 }, { "epoch": 5.71, "learning_rate": 2.6477295459091823e-06, "loss": 0.0094, "step": 18389 }, { "epoch": 5.71, "learning_rate": 2.6473294658931786e-06, "loss": 0.0023, "step": 18390 }, { "epoch": 5.71, "learning_rate": 2.6469293858771757e-06, "loss": 0.0048, "step": 18391 }, { "epoch": 5.71, "learning_rate": 2.6465293058611724e-06, "loss": 0.0046, "step": 18392 }, { "epoch": 5.71, "learning_rate": 2.6461292258451695e-06, "loss": 0.0054, "step": 18393 }, { "epoch": 5.71, "learning_rate": 2.645729145829166e-06, "loss": 0.0054, "step": 18394 }, { "epoch": 5.71, "learning_rate": 2.6453290658131633e-06, "loss": 0.0033, "step": 18395 }, { "epoch": 5.71, "learning_rate": 2.6449289857971595e-06, "loss": 0.0105, "step": 18396 }, { "epoch": 5.71, "learning_rate": 2.644528905781156e-06, "loss": 0.0291, "step": 18397 }, { "epoch": 5.71, "learning_rate": 2.6441288257651533e-06, "loss": 0.0284, "step": 18398 }, { "epoch": 5.71, "learning_rate": 2.64372874574915e-06, "loss": 0.0032, "step": 18399 }, { "epoch": 5.71, "learning_rate": 2.643328665733147e-06, "loss": 0.0154, "step": 18400 }, { "epoch": 5.71, "learning_rate": 2.642928585717144e-06, "loss": 0.0057, "step": 18401 }, { "epoch": 5.71, "learning_rate": 2.64252850570114e-06, "loss": 0.0151, "step": 18402 }, { "epoch": 5.71, "learning_rate": 2.642128425685137e-06, "loss": 0.0033, "step": 18403 }, { "epoch": 5.71, "learning_rate": 2.641728345669134e-06, "loss": 0.0488, "step": 18404 }, { "epoch": 5.71, "learning_rate": 2.641328265653131e-06, "loss": 0.0044, "step": 18405 }, { "epoch": 5.71, "learning_rate": 2.6409281856371277e-06, "loss": 0.0134, "step": 18406 }, { "epoch": 5.71, "learning_rate": 2.6405281056211248e-06, "loss": 0.0071, "step": 18407 }, { "epoch": 5.71, "learning_rate": 2.6401280256051214e-06, "loss": 0.0071, "step": 18408 }, { "epoch": 5.71, "learning_rate": 2.6397279455891177e-06, "loss": 0.0023, "step": 18409 }, { "epoch": 5.71, "learning_rate": 2.639327865573115e-06, "loss": 0.0034, "step": 18410 }, { "epoch": 5.71, "learning_rate": 2.6389277855571115e-06, "loss": 0.049, "step": 18411 }, { "epoch": 5.71, "learning_rate": 2.6385277055411086e-06, "loss": 0.0053, "step": 18412 }, { "epoch": 5.71, "learning_rate": 2.6381276255251053e-06, "loss": 0.0226, "step": 18413 }, { "epoch": 5.72, "learning_rate": 2.6377275455091024e-06, "loss": 0.0027, "step": 18414 }, { "epoch": 5.72, "learning_rate": 2.6373274654930987e-06, "loss": 0.0083, "step": 18415 }, { "epoch": 5.72, "learning_rate": 2.6369273854770954e-06, "loss": 0.0039, "step": 18416 }, { "epoch": 5.72, "learning_rate": 2.6365273054610925e-06, "loss": 0.0443, "step": 18417 }, { "epoch": 5.72, "learning_rate": 2.636127225445089e-06, "loss": 0.0394, "step": 18418 }, { "epoch": 5.72, "learning_rate": 2.6357271454290863e-06, "loss": 0.0179, "step": 18419 }, { "epoch": 5.72, "learning_rate": 2.635327065413083e-06, "loss": 0.0506, "step": 18420 }, { "epoch": 5.72, "learning_rate": 2.6349269853970792e-06, "loss": 0.0113, "step": 18421 }, { "epoch": 5.72, "learning_rate": 2.6345269053810763e-06, "loss": 0.0037, "step": 18422 }, { "epoch": 5.72, "learning_rate": 2.634126825365073e-06, "loss": 0.0069, "step": 18423 }, { "epoch": 5.72, "learning_rate": 2.63372674534907e-06, "loss": 0.0029, "step": 18424 }, { "epoch": 5.72, "learning_rate": 2.633326665333067e-06, "loss": 0.0043, "step": 18425 }, { "epoch": 5.72, "learning_rate": 2.632926585317064e-06, "loss": 0.0017, "step": 18426 }, { "epoch": 5.72, "learning_rate": 2.6325265053010606e-06, "loss": 0.0032, "step": 18427 }, { "epoch": 5.72, "learning_rate": 2.632126425285057e-06, "loss": 0.0034, "step": 18428 }, { "epoch": 5.72, "learning_rate": 2.631726345269054e-06, "loss": 0.0076, "step": 18429 }, { "epoch": 5.72, "learning_rate": 2.6313262652530507e-06, "loss": 0.0035, "step": 18430 }, { "epoch": 5.72, "learning_rate": 2.6309261852370478e-06, "loss": 0.0445, "step": 18431 }, { "epoch": 5.72, "learning_rate": 2.6305261052210445e-06, "loss": 0.0085, "step": 18432 }, { "epoch": 5.72, "learning_rate": 2.6301260252050416e-06, "loss": 0.0057, "step": 18433 }, { "epoch": 5.72, "learning_rate": 2.629725945189038e-06, "loss": 0.0076, "step": 18434 }, { "epoch": 5.72, "learning_rate": 2.6293258651730345e-06, "loss": 0.0139, "step": 18435 }, { "epoch": 5.72, "learning_rate": 2.6289257851570316e-06, "loss": 0.0138, "step": 18436 }, { "epoch": 5.72, "learning_rate": 2.6285257051410283e-06, "loss": 0.0026, "step": 18437 }, { "epoch": 5.72, "learning_rate": 2.6281256251250254e-06, "loss": 0.0087, "step": 18438 }, { "epoch": 5.72, "learning_rate": 2.627725545109022e-06, "loss": 0.0015, "step": 18439 }, { "epoch": 5.72, "learning_rate": 2.6273254650930184e-06, "loss": 0.0078, "step": 18440 }, { "epoch": 5.72, "learning_rate": 2.6269253850770155e-06, "loss": 0.0029, "step": 18441 }, { "epoch": 5.72, "learning_rate": 2.626525305061012e-06, "loss": 0.0029, "step": 18442 }, { "epoch": 5.72, "learning_rate": 2.6261252250450093e-06, "loss": 0.0176, "step": 18443 }, { "epoch": 5.72, "learning_rate": 2.625725145029006e-06, "loss": 0.004, "step": 18444 }, { "epoch": 5.72, "learning_rate": 2.625325065013003e-06, "loss": 0.0054, "step": 18445 }, { "epoch": 5.73, "learning_rate": 2.6249249849969998e-06, "loss": 0.0068, "step": 18446 }, { "epoch": 5.73, "learning_rate": 2.624524904980996e-06, "loss": 0.002, "step": 18447 }, { "epoch": 5.73, "learning_rate": 2.624124824964993e-06, "loss": 0.0049, "step": 18448 }, { "epoch": 5.73, "learning_rate": 2.62372474494899e-06, "loss": 0.0071, "step": 18449 }, { "epoch": 5.73, "learning_rate": 2.623324664932987e-06, "loss": 0.0079, "step": 18450 }, { "epoch": 5.73, "learning_rate": 2.6229245849169836e-06, "loss": 0.0034, "step": 18451 }, { "epoch": 5.73, "learning_rate": 2.6225245049009807e-06, "loss": 0.0124, "step": 18452 }, { "epoch": 5.73, "learning_rate": 2.622124424884977e-06, "loss": 0.0047, "step": 18453 }, { "epoch": 5.73, "learning_rate": 2.6217243448689737e-06, "loss": 0.003, "step": 18454 }, { "epoch": 5.73, "learning_rate": 2.6213242648529708e-06, "loss": 0.0102, "step": 18455 }, { "epoch": 5.73, "learning_rate": 2.6209241848369675e-06, "loss": 0.0037, "step": 18456 }, { "epoch": 5.73, "learning_rate": 2.6205241048209646e-06, "loss": 0.0056, "step": 18457 }, { "epoch": 5.73, "learning_rate": 2.6201240248049613e-06, "loss": 0.0025, "step": 18458 }, { "epoch": 5.73, "learning_rate": 2.619723944788958e-06, "loss": 0.0044, "step": 18459 }, { "epoch": 5.73, "learning_rate": 2.6193238647729546e-06, "loss": 0.0133, "step": 18460 }, { "epoch": 5.73, "learning_rate": 2.6189237847569513e-06, "loss": 0.0065, "step": 18461 }, { "epoch": 5.73, "learning_rate": 2.6185237047409484e-06, "loss": 0.0081, "step": 18462 }, { "epoch": 5.73, "learning_rate": 2.618123624724945e-06, "loss": 0.0519, "step": 18463 }, { "epoch": 5.73, "learning_rate": 2.6177235447089422e-06, "loss": 0.0024, "step": 18464 }, { "epoch": 5.73, "learning_rate": 2.617323464692939e-06, "loss": 0.0485, "step": 18465 }, { "epoch": 5.73, "learning_rate": 2.6169233846769356e-06, "loss": 0.0248, "step": 18466 }, { "epoch": 5.73, "learning_rate": 2.6165233046609323e-06, "loss": 0.0037, "step": 18467 }, { "epoch": 5.73, "learning_rate": 2.616123224644929e-06, "loss": 0.0179, "step": 18468 }, { "epoch": 5.73, "learning_rate": 2.615723144628926e-06, "loss": 0.002, "step": 18469 }, { "epoch": 5.73, "learning_rate": 2.6153230646129228e-06, "loss": 0.0014, "step": 18470 }, { "epoch": 5.73, "learning_rate": 2.61492298459692e-06, "loss": 0.0166, "step": 18471 }, { "epoch": 5.73, "learning_rate": 2.614522904580916e-06, "loss": 0.0098, "step": 18472 }, { "epoch": 5.73, "learning_rate": 2.6141228245649133e-06, "loss": 0.0763, "step": 18473 }, { "epoch": 5.73, "learning_rate": 2.61372274454891e-06, "loss": 0.0096, "step": 18474 }, { "epoch": 5.73, "learning_rate": 2.6133226645329066e-06, "loss": 0.0596, "step": 18475 }, { "epoch": 5.73, "learning_rate": 2.6129225845169037e-06, "loss": 0.006, "step": 18476 }, { "epoch": 5.73, "learning_rate": 2.6125225045009004e-06, "loss": 0.0025, "step": 18477 }, { "epoch": 5.74, "learning_rate": 2.612122424484897e-06, "loss": 0.0055, "step": 18478 }, { "epoch": 5.74, "learning_rate": 2.611722344468894e-06, "loss": 0.0064, "step": 18479 }, { "epoch": 5.74, "learning_rate": 2.611322264452891e-06, "loss": 0.0012, "step": 18480 }, { "epoch": 5.74, "learning_rate": 2.6109221844368876e-06, "loss": 0.0008, "step": 18481 }, { "epoch": 5.74, "learning_rate": 2.6105221044208847e-06, "loss": 0.0038, "step": 18482 }, { "epoch": 5.74, "learning_rate": 2.6101220244048814e-06, "loss": 0.0029, "step": 18483 }, { "epoch": 5.74, "learning_rate": 2.6097219443888777e-06, "loss": 0.0681, "step": 18484 }, { "epoch": 5.74, "learning_rate": 2.6093218643728748e-06, "loss": 0.011, "step": 18485 }, { "epoch": 5.74, "learning_rate": 2.6089217843568714e-06, "loss": 0.0049, "step": 18486 }, { "epoch": 5.74, "learning_rate": 2.6085217043408686e-06, "loss": 0.0018, "step": 18487 }, { "epoch": 5.74, "learning_rate": 2.6081216243248652e-06, "loss": 0.0078, "step": 18488 }, { "epoch": 5.74, "learning_rate": 2.6077215443088624e-06, "loss": 0.0169, "step": 18489 }, { "epoch": 5.74, "learning_rate": 2.607321464292859e-06, "loss": 0.0038, "step": 18490 }, { "epoch": 5.74, "learning_rate": 2.6069213842768553e-06, "loss": 0.0097, "step": 18491 }, { "epoch": 5.74, "learning_rate": 2.6065213042608524e-06, "loss": 0.0048, "step": 18492 }, { "epoch": 5.74, "learning_rate": 2.606121224244849e-06, "loss": 0.0312, "step": 18493 }, { "epoch": 5.74, "learning_rate": 2.605721144228846e-06, "loss": 0.0236, "step": 18494 }, { "epoch": 5.74, "learning_rate": 2.605321064212843e-06, "loss": 0.0136, "step": 18495 }, { "epoch": 5.74, "learning_rate": 2.60492098419684e-06, "loss": 0.0024, "step": 18496 }, { "epoch": 5.74, "learning_rate": 2.6045209041808363e-06, "loss": 0.0072, "step": 18497 }, { "epoch": 5.74, "learning_rate": 2.604120824164833e-06, "loss": 0.004, "step": 18498 }, { "epoch": 5.74, "learning_rate": 2.60372074414883e-06, "loss": 0.0048, "step": 18499 }, { "epoch": 5.74, "learning_rate": 2.6033206641328268e-06, "loss": 0.0068, "step": 18500 }, { "epoch": 5.74, "learning_rate": 2.602920584116824e-06, "loss": 0.0248, "step": 18501 }, { "epoch": 5.74, "learning_rate": 2.6025205041008205e-06, "loss": 0.0027, "step": 18502 }, { "epoch": 5.74, "learning_rate": 2.602120424084817e-06, "loss": 0.0304, "step": 18503 }, { "epoch": 5.74, "learning_rate": 2.601720344068814e-06, "loss": 0.0036, "step": 18504 }, { "epoch": 5.74, "learning_rate": 2.6013202640528106e-06, "loss": 0.0051, "step": 18505 }, { "epoch": 5.74, "learning_rate": 2.6009201840368077e-06, "loss": 0.0208, "step": 18506 }, { "epoch": 5.74, "learning_rate": 2.6005201040208044e-06, "loss": 0.0023, "step": 18507 }, { "epoch": 5.74, "learning_rate": 2.6001200240048015e-06, "loss": 0.0041, "step": 18508 }, { "epoch": 5.74, "learning_rate": 2.599719943988798e-06, "loss": 0.0139, "step": 18509 }, { "epoch": 5.75, "learning_rate": 2.5993198639727945e-06, "loss": 0.0139, "step": 18510 }, { "epoch": 5.75, "learning_rate": 2.5989197839567916e-06, "loss": 0.0136, "step": 18511 }, { "epoch": 5.75, "learning_rate": 2.5985197039407883e-06, "loss": 0.0184, "step": 18512 }, { "epoch": 5.75, "learning_rate": 2.5981196239247854e-06, "loss": 0.0085, "step": 18513 }, { "epoch": 5.75, "learning_rate": 2.597719543908782e-06, "loss": 0.0073, "step": 18514 }, { "epoch": 5.75, "learning_rate": 2.597319463892779e-06, "loss": 0.0016, "step": 18515 }, { "epoch": 5.75, "learning_rate": 2.5969193838767754e-06, "loss": 0.0131, "step": 18516 }, { "epoch": 5.75, "learning_rate": 2.596519303860772e-06, "loss": 0.0028, "step": 18517 }, { "epoch": 5.75, "learning_rate": 2.5961192238447692e-06, "loss": 0.0128, "step": 18518 }, { "epoch": 5.75, "learning_rate": 2.595719143828766e-06, "loss": 0.001, "step": 18519 }, { "epoch": 5.75, "learning_rate": 2.595319063812763e-06, "loss": 0.0047, "step": 18520 }, { "epoch": 5.75, "learning_rate": 2.5949189837967597e-06, "loss": 0.0048, "step": 18521 }, { "epoch": 5.75, "learning_rate": 2.594518903780756e-06, "loss": 0.004, "step": 18522 }, { "epoch": 5.75, "learning_rate": 2.594118823764753e-06, "loss": 0.0657, "step": 18523 }, { "epoch": 5.75, "learning_rate": 2.5937187437487498e-06, "loss": 0.0428, "step": 18524 }, { "epoch": 5.75, "learning_rate": 2.593318663732747e-06, "loss": 0.0155, "step": 18525 }, { "epoch": 5.75, "learning_rate": 2.5929185837167436e-06, "loss": 0.0059, "step": 18526 }, { "epoch": 5.75, "learning_rate": 2.5925185037007407e-06, "loss": 0.0054, "step": 18527 }, { "epoch": 5.75, "learning_rate": 2.5921184236847374e-06, "loss": 0.008, "step": 18528 }, { "epoch": 5.75, "learning_rate": 2.5917183436687336e-06, "loss": 0.0314, "step": 18529 }, { "epoch": 5.75, "learning_rate": 2.5913182636527307e-06, "loss": 0.0037, "step": 18530 }, { "epoch": 5.75, "learning_rate": 2.5909181836367274e-06, "loss": 0.0175, "step": 18531 }, { "epoch": 5.75, "learning_rate": 2.5905181036207245e-06, "loss": 0.0182, "step": 18532 }, { "epoch": 5.75, "learning_rate": 2.590118023604721e-06, "loss": 0.0026, "step": 18533 }, { "epoch": 5.75, "learning_rate": 2.5897179435887183e-06, "loss": 0.0034, "step": 18534 }, { "epoch": 5.75, "learning_rate": 2.5893178635727146e-06, "loss": 0.003, "step": 18535 }, { "epoch": 5.75, "learning_rate": 2.5889177835567113e-06, "loss": 0.0743, "step": 18536 }, { "epoch": 5.75, "learning_rate": 2.5885177035407084e-06, "loss": 0.0035, "step": 18537 }, { "epoch": 5.75, "learning_rate": 2.588117623524705e-06, "loss": 0.0251, "step": 18538 }, { "epoch": 5.75, "learning_rate": 2.587717543508702e-06, "loss": 0.0116, "step": 18539 }, { "epoch": 5.75, "learning_rate": 2.587317463492699e-06, "loss": 0.0041, "step": 18540 }, { "epoch": 5.75, "learning_rate": 2.586917383476695e-06, "loss": 0.0087, "step": 18541 }, { "epoch": 5.76, "learning_rate": 2.5865173034606922e-06, "loss": 0.0135, "step": 18542 }, { "epoch": 5.76, "learning_rate": 2.586117223444689e-06, "loss": 0.0347, "step": 18543 }, { "epoch": 5.76, "learning_rate": 2.585717143428686e-06, "loss": 0.0078, "step": 18544 }, { "epoch": 5.76, "learning_rate": 2.5853170634126827e-06, "loss": 0.0026, "step": 18545 }, { "epoch": 5.76, "learning_rate": 2.58491698339668e-06, "loss": 0.002, "step": 18546 }, { "epoch": 5.76, "learning_rate": 2.5845169033806765e-06, "loss": 0.0062, "step": 18547 }, { "epoch": 5.76, "learning_rate": 2.5841168233646728e-06, "loss": 0.0133, "step": 18548 }, { "epoch": 5.76, "learning_rate": 2.58371674334867e-06, "loss": 0.0048, "step": 18549 }, { "epoch": 5.76, "learning_rate": 2.5833166633326666e-06, "loss": 0.017, "step": 18550 }, { "epoch": 5.76, "learning_rate": 2.5829165833166637e-06, "loss": 0.05, "step": 18551 }, { "epoch": 5.76, "learning_rate": 2.5825165033006604e-06, "loss": 0.0151, "step": 18552 }, { "epoch": 5.76, "learning_rate": 2.5821164232846575e-06, "loss": 0.065, "step": 18553 }, { "epoch": 5.76, "learning_rate": 2.5817163432686537e-06, "loss": 0.0019, "step": 18554 }, { "epoch": 5.76, "learning_rate": 2.5813162632526504e-06, "loss": 0.0259, "step": 18555 }, { "epoch": 5.76, "learning_rate": 2.5809161832366475e-06, "loss": 0.0091, "step": 18556 }, { "epoch": 5.76, "learning_rate": 2.5805161032206442e-06, "loss": 0.0042, "step": 18557 }, { "epoch": 5.76, "learning_rate": 2.5801160232046413e-06, "loss": 0.0094, "step": 18558 }, { "epoch": 5.76, "learning_rate": 2.579715943188638e-06, "loss": 0.007, "step": 18559 }, { "epoch": 5.76, "learning_rate": 2.5793158631726347e-06, "loss": 0.0026, "step": 18560 }, { "epoch": 5.76, "learning_rate": 2.5789157831566314e-06, "loss": 0.0024, "step": 18561 }, { "epoch": 5.76, "learning_rate": 2.578515703140628e-06, "loss": 0.0065, "step": 18562 }, { "epoch": 5.76, "learning_rate": 2.578115623124625e-06, "loss": 0.0068, "step": 18563 }, { "epoch": 5.76, "learning_rate": 2.577715543108622e-06, "loss": 0.0087, "step": 18564 }, { "epoch": 5.76, "learning_rate": 2.577315463092619e-06, "loss": 0.0075, "step": 18565 }, { "epoch": 5.76, "learning_rate": 2.5769153830766152e-06, "loss": 0.0327, "step": 18566 }, { "epoch": 5.76, "learning_rate": 2.5765153030606124e-06, "loss": 0.0528, "step": 18567 }, { "epoch": 5.76, "learning_rate": 2.576115223044609e-06, "loss": 0.0079, "step": 18568 }, { "epoch": 5.76, "learning_rate": 2.5757151430286057e-06, "loss": 0.0075, "step": 18569 }, { "epoch": 5.76, "learning_rate": 2.575315063012603e-06, "loss": 0.003, "step": 18570 }, { "epoch": 5.76, "learning_rate": 2.5749149829965995e-06, "loss": 0.0089, "step": 18571 }, { "epoch": 5.76, "learning_rate": 2.5745149029805966e-06, "loss": 0.0526, "step": 18572 }, { "epoch": 5.76, "learning_rate": 2.574114822964593e-06, "loss": 0.0501, "step": 18573 }, { "epoch": 5.76, "learning_rate": 2.57371474294859e-06, "loss": 0.0015, "step": 18574 }, { "epoch": 5.77, "learning_rate": 2.5733146629325867e-06, "loss": 0.0127, "step": 18575 }, { "epoch": 5.77, "learning_rate": 2.5729145829165834e-06, "loss": 0.0039, "step": 18576 }, { "epoch": 5.77, "learning_rate": 2.5725145029005805e-06, "loss": 0.0024, "step": 18577 }, { "epoch": 5.77, "learning_rate": 2.572114422884577e-06, "loss": 0.0074, "step": 18578 }, { "epoch": 5.77, "learning_rate": 2.571714342868574e-06, "loss": 0.0136, "step": 18579 }, { "epoch": 5.77, "learning_rate": 2.5713142628525705e-06, "loss": 0.0203, "step": 18580 }, { "epoch": 5.77, "learning_rate": 2.5709141828365677e-06, "loss": 0.0041, "step": 18581 }, { "epoch": 5.77, "learning_rate": 2.5705141028205643e-06, "loss": 0.0251, "step": 18582 }, { "epoch": 5.77, "learning_rate": 2.570114022804561e-06, "loss": 0.008, "step": 18583 }, { "epoch": 5.77, "learning_rate": 2.569713942788558e-06, "loss": 0.0028, "step": 18584 }, { "epoch": 5.77, "learning_rate": 2.5693138627725544e-06, "loss": 0.0052, "step": 18585 }, { "epoch": 5.77, "learning_rate": 2.5689137827565515e-06, "loss": 0.0057, "step": 18586 }, { "epoch": 5.77, "learning_rate": 2.568513702740548e-06, "loss": 0.0031, "step": 18587 }, { "epoch": 5.77, "learning_rate": 2.5681136227245453e-06, "loss": 0.0067, "step": 18588 }, { "epoch": 5.77, "learning_rate": 2.567713542708542e-06, "loss": 0.012, "step": 18589 }, { "epoch": 5.77, "learning_rate": 2.567313462692539e-06, "loss": 0.0051, "step": 18590 }, { "epoch": 5.77, "learning_rate": 2.566913382676536e-06, "loss": 0.0026, "step": 18591 }, { "epoch": 5.77, "learning_rate": 2.566513302660532e-06, "loss": 0.0075, "step": 18592 }, { "epoch": 5.77, "learning_rate": 2.566113222644529e-06, "loss": 0.0012, "step": 18593 }, { "epoch": 5.77, "learning_rate": 2.565713142628526e-06, "loss": 0.0127, "step": 18594 }, { "epoch": 5.77, "learning_rate": 2.565313062612523e-06, "loss": 0.0043, "step": 18595 }, { "epoch": 5.77, "learning_rate": 2.5649129825965196e-06, "loss": 0.0047, "step": 18596 }, { "epoch": 5.77, "learning_rate": 2.5645129025805168e-06, "loss": 0.0051, "step": 18597 }, { "epoch": 5.77, "learning_rate": 2.564112822564513e-06, "loss": 0.0099, "step": 18598 }, { "epoch": 5.77, "learning_rate": 2.5637127425485097e-06, "loss": 0.0036, "step": 18599 }, { "epoch": 5.77, "learning_rate": 2.563312662532507e-06, "loss": 0.0069, "step": 18600 }, { "epoch": 5.77, "learning_rate": 2.5629125825165035e-06, "loss": 0.051, "step": 18601 }, { "epoch": 5.77, "learning_rate": 2.5625125025005006e-06, "loss": 0.0085, "step": 18602 }, { "epoch": 5.77, "learning_rate": 2.5621124224844973e-06, "loss": 0.0315, "step": 18603 }, { "epoch": 5.77, "learning_rate": 2.5617123424684936e-06, "loss": 0.0044, "step": 18604 }, { "epoch": 5.77, "learning_rate": 2.5613122624524907e-06, "loss": 0.0119, "step": 18605 }, { "epoch": 5.77, "learning_rate": 2.5609121824364874e-06, "loss": 0.0167, "step": 18606 }, { "epoch": 5.78, "learning_rate": 2.5605121024204845e-06, "loss": 0.0158, "step": 18607 }, { "epoch": 5.78, "learning_rate": 2.560112022404481e-06, "loss": 0.0025, "step": 18608 }, { "epoch": 5.78, "learning_rate": 2.5597119423884783e-06, "loss": 0.0144, "step": 18609 }, { "epoch": 5.78, "learning_rate": 2.559311862372475e-06, "loss": 0.0035, "step": 18610 }, { "epoch": 5.78, "learning_rate": 2.558911782356471e-06, "loss": 0.0062, "step": 18611 }, { "epoch": 5.78, "learning_rate": 2.5585117023404683e-06, "loss": 0.0222, "step": 18612 }, { "epoch": 5.78, "learning_rate": 2.558111622324465e-06, "loss": 0.001, "step": 18613 }, { "epoch": 5.78, "learning_rate": 2.557711542308462e-06, "loss": 0.0084, "step": 18614 }, { "epoch": 5.78, "learning_rate": 2.557311462292459e-06, "loss": 0.0097, "step": 18615 }, { "epoch": 5.78, "learning_rate": 2.556911382276456e-06, "loss": 0.0087, "step": 18616 }, { "epoch": 5.78, "learning_rate": 2.556511302260452e-06, "loss": 0.0095, "step": 18617 }, { "epoch": 5.78, "learning_rate": 2.556111222244449e-06, "loss": 0.0053, "step": 18618 }, { "epoch": 5.78, "learning_rate": 2.555711142228446e-06, "loss": 0.0055, "step": 18619 }, { "epoch": 5.78, "learning_rate": 2.5553110622124427e-06, "loss": 0.0025, "step": 18620 }, { "epoch": 5.78, "learning_rate": 2.5549109821964398e-06, "loss": 0.0039, "step": 18621 }, { "epoch": 5.78, "learning_rate": 2.5545109021804365e-06, "loss": 0.0073, "step": 18622 }, { "epoch": 5.78, "learning_rate": 2.5541108221644327e-06, "loss": 0.002, "step": 18623 }, { "epoch": 5.78, "learning_rate": 2.55371074214843e-06, "loss": 0.0047, "step": 18624 }, { "epoch": 5.78, "learning_rate": 2.5533106621324265e-06, "loss": 0.0101, "step": 18625 }, { "epoch": 5.78, "learning_rate": 2.5529105821164236e-06, "loss": 0.0106, "step": 18626 }, { "epoch": 5.78, "learning_rate": 2.5525105021004203e-06, "loss": 0.0335, "step": 18627 }, { "epoch": 5.78, "learning_rate": 2.5521104220844174e-06, "loss": 0.0046, "step": 18628 }, { "epoch": 5.78, "learning_rate": 2.5517103420684137e-06, "loss": 0.0062, "step": 18629 }, { "epoch": 5.78, "learning_rate": 2.5513102620524104e-06, "loss": 0.0422, "step": 18630 }, { "epoch": 5.78, "learning_rate": 2.5509101820364075e-06, "loss": 0.0025, "step": 18631 }, { "epoch": 5.78, "learning_rate": 2.550510102020404e-06, "loss": 0.0145, "step": 18632 }, { "epoch": 5.78, "learning_rate": 2.5501100220044013e-06, "loss": 0.0096, "step": 18633 }, { "epoch": 5.78, "learning_rate": 2.549709941988398e-06, "loss": 0.0113, "step": 18634 }, { "epoch": 5.78, "learning_rate": 2.549309861972395e-06, "loss": 0.0067, "step": 18635 }, { "epoch": 5.78, "learning_rate": 2.5489097819563913e-06, "loss": 0.006, "step": 18636 }, { "epoch": 5.78, "learning_rate": 2.548509701940388e-06, "loss": 0.0095, "step": 18637 }, { "epoch": 5.78, "learning_rate": 2.548109621924385e-06, "loss": 0.0053, "step": 18638 }, { "epoch": 5.79, "learning_rate": 2.547709541908382e-06, "loss": 0.0051, "step": 18639 }, { "epoch": 5.79, "learning_rate": 2.547309461892379e-06, "loss": 0.0078, "step": 18640 }, { "epoch": 5.79, "learning_rate": 2.5469093818763756e-06, "loss": 0.006, "step": 18641 }, { "epoch": 5.79, "learning_rate": 2.546509301860372e-06, "loss": 0.0046, "step": 18642 }, { "epoch": 5.79, "learning_rate": 2.546109221844369e-06, "loss": 0.0176, "step": 18643 }, { "epoch": 5.79, "learning_rate": 2.5457091418283657e-06, "loss": 0.006, "step": 18644 }, { "epoch": 5.79, "learning_rate": 2.5453090618123628e-06, "loss": 0.0019, "step": 18645 }, { "epoch": 5.79, "learning_rate": 2.5449089817963595e-06, "loss": 0.0014, "step": 18646 }, { "epoch": 5.79, "learning_rate": 2.5445089017803566e-06, "loss": 0.0085, "step": 18647 }, { "epoch": 5.79, "learning_rate": 2.544108821764353e-06, "loss": 0.0077, "step": 18648 }, { "epoch": 5.79, "learning_rate": 2.5437087417483495e-06, "loss": 0.0061, "step": 18649 }, { "epoch": 5.79, "learning_rate": 2.5433086617323466e-06, "loss": 0.0021, "step": 18650 }, { "epoch": 5.79, "learning_rate": 2.5429085817163433e-06, "loss": 0.0045, "step": 18651 }, { "epoch": 5.79, "learning_rate": 2.5425085017003404e-06, "loss": 0.0055, "step": 18652 }, { "epoch": 5.79, "learning_rate": 2.542108421684337e-06, "loss": 0.007, "step": 18653 }, { "epoch": 5.79, "learning_rate": 2.5417083416683342e-06, "loss": 0.0055, "step": 18654 }, { "epoch": 5.79, "learning_rate": 2.5413082616523305e-06, "loss": 0.004, "step": 18655 }, { "epoch": 5.79, "learning_rate": 2.540908181636327e-06, "loss": 0.0069, "step": 18656 }, { "epoch": 5.79, "learning_rate": 2.5405081016203243e-06, "loss": 0.0154, "step": 18657 }, { "epoch": 5.79, "learning_rate": 2.540108021604321e-06, "loss": 0.0039, "step": 18658 }, { "epoch": 5.79, "learning_rate": 2.539707941588318e-06, "loss": 0.0028, "step": 18659 }, { "epoch": 5.79, "learning_rate": 2.5393078615723148e-06, "loss": 0.0037, "step": 18660 }, { "epoch": 5.79, "learning_rate": 2.538907781556311e-06, "loss": 0.0033, "step": 18661 }, { "epoch": 5.79, "learning_rate": 2.538507701540308e-06, "loss": 0.0095, "step": 18662 }, { "epoch": 5.79, "learning_rate": 2.538107621524305e-06, "loss": 0.0038, "step": 18663 }, { "epoch": 5.79, "learning_rate": 2.537707541508302e-06, "loss": 0.0624, "step": 18664 }, { "epoch": 5.79, "learning_rate": 2.5373074614922986e-06, "loss": 0.0039, "step": 18665 }, { "epoch": 5.79, "learning_rate": 2.5369073814762957e-06, "loss": 0.0032, "step": 18666 }, { "epoch": 5.79, "learning_rate": 2.536507301460292e-06, "loss": 0.0131, "step": 18667 }, { "epoch": 5.79, "learning_rate": 2.5361072214442887e-06, "loss": 0.0047, "step": 18668 }, { "epoch": 5.79, "learning_rate": 2.535707141428286e-06, "loss": 0.0105, "step": 18669 }, { "epoch": 5.79, "learning_rate": 2.5353070614122825e-06, "loss": 0.0066, "step": 18670 }, { "epoch": 5.8, "learning_rate": 2.5349069813962796e-06, "loss": 0.0262, "step": 18671 }, { "epoch": 5.8, "learning_rate": 2.5345069013802763e-06, "loss": 0.0098, "step": 18672 }, { "epoch": 5.8, "learning_rate": 2.5341068213642734e-06, "loss": 0.0279, "step": 18673 }, { "epoch": 5.8, "learning_rate": 2.5337067413482696e-06, "loss": 0.0038, "step": 18674 }, { "epoch": 5.8, "learning_rate": 2.5333066613322668e-06, "loss": 0.017, "step": 18675 }, { "epoch": 5.8, "learning_rate": 2.5329065813162634e-06, "loss": 0.0021, "step": 18676 }, { "epoch": 5.8, "learning_rate": 2.53250650130026e-06, "loss": 0.004, "step": 18677 }, { "epoch": 5.8, "learning_rate": 2.5321064212842572e-06, "loss": 0.0021, "step": 18678 }, { "epoch": 5.8, "learning_rate": 2.531706341268254e-06, "loss": 0.0042, "step": 18679 }, { "epoch": 5.8, "learning_rate": 2.5313062612522506e-06, "loss": 0.0138, "step": 18680 }, { "epoch": 5.8, "learning_rate": 2.5309061812362473e-06, "loss": 0.0045, "step": 18681 }, { "epoch": 5.8, "learning_rate": 2.5305061012202444e-06, "loss": 0.0478, "step": 18682 }, { "epoch": 5.8, "learning_rate": 2.530106021204241e-06, "loss": 0.0017, "step": 18683 }, { "epoch": 5.8, "learning_rate": 2.5297059411882378e-06, "loss": 0.0025, "step": 18684 }, { "epoch": 5.8, "learning_rate": 2.529305861172235e-06, "loss": 0.0024, "step": 18685 }, { "epoch": 5.8, "learning_rate": 2.528905781156231e-06, "loss": 0.0021, "step": 18686 }, { "epoch": 5.8, "learning_rate": 2.5285057011402283e-06, "loss": 0.0081, "step": 18687 }, { "epoch": 5.8, "learning_rate": 2.528105621124225e-06, "loss": 0.0038, "step": 18688 }, { "epoch": 5.8, "learning_rate": 2.527705541108222e-06, "loss": 0.0048, "step": 18689 }, { "epoch": 5.8, "learning_rate": 2.5273054610922187e-06, "loss": 0.0099, "step": 18690 }, { "epoch": 5.8, "learning_rate": 2.5269053810762154e-06, "loss": 0.0036, "step": 18691 }, { "epoch": 5.8, "learning_rate": 2.5265053010602125e-06, "loss": 0.0163, "step": 18692 }, { "epoch": 5.8, "learning_rate": 2.526105221044209e-06, "loss": 0.0031, "step": 18693 }, { "epoch": 5.8, "learning_rate": 2.525705141028206e-06, "loss": 0.0108, "step": 18694 }, { "epoch": 5.8, "learning_rate": 2.5253050610122026e-06, "loss": 0.0016, "step": 18695 }, { "epoch": 5.8, "learning_rate": 2.5249049809961997e-06, "loss": 0.0079, "step": 18696 }, { "epoch": 5.8, "learning_rate": 2.5245049009801964e-06, "loss": 0.0082, "step": 18697 }, { "epoch": 5.8, "learning_rate": 2.5241048209641935e-06, "loss": 0.0316, "step": 18698 }, { "epoch": 5.8, "learning_rate": 2.5237047409481898e-06, "loss": 0.0071, "step": 18699 }, { "epoch": 5.8, "learning_rate": 2.5233046609321865e-06, "loss": 0.0032, "step": 18700 }, { "epoch": 5.8, "learning_rate": 2.5229045809161836e-06, "loss": 0.0057, "step": 18701 }, { "epoch": 5.8, "learning_rate": 2.5225045009001802e-06, "loss": 0.002, "step": 18702 }, { "epoch": 5.81, "learning_rate": 2.5221044208841774e-06, "loss": 0.0036, "step": 18703 }, { "epoch": 5.81, "learning_rate": 2.521704340868174e-06, "loss": 0.0872, "step": 18704 }, { "epoch": 5.81, "learning_rate": 2.5213042608521703e-06, "loss": 0.0072, "step": 18705 }, { "epoch": 5.81, "learning_rate": 2.5209041808361674e-06, "loss": 0.0113, "step": 18706 }, { "epoch": 5.81, "learning_rate": 2.520504100820164e-06, "loss": 0.016, "step": 18707 }, { "epoch": 5.81, "learning_rate": 2.5201040208041612e-06, "loss": 0.004, "step": 18708 }, { "epoch": 5.81, "learning_rate": 2.519703940788158e-06, "loss": 0.0058, "step": 18709 }, { "epoch": 5.81, "learning_rate": 2.519303860772155e-06, "loss": 0.0084, "step": 18710 }, { "epoch": 5.81, "learning_rate": 2.5189037807561513e-06, "loss": 0.0099, "step": 18711 }, { "epoch": 5.81, "learning_rate": 2.518503700740148e-06, "loss": 0.0092, "step": 18712 }, { "epoch": 5.81, "learning_rate": 2.518103620724145e-06, "loss": 0.0172, "step": 18713 }, { "epoch": 5.81, "learning_rate": 2.5177035407081418e-06, "loss": 0.007, "step": 18714 }, { "epoch": 5.81, "learning_rate": 2.517303460692139e-06, "loss": 0.0312, "step": 18715 }, { "epoch": 5.81, "learning_rate": 2.5169033806761356e-06, "loss": 0.0023, "step": 18716 }, { "epoch": 5.81, "learning_rate": 2.5165033006601327e-06, "loss": 0.0455, "step": 18717 }, { "epoch": 5.81, "learning_rate": 2.516103220644129e-06, "loss": 0.0029, "step": 18718 }, { "epoch": 5.81, "learning_rate": 2.5157031406281256e-06, "loss": 0.0042, "step": 18719 }, { "epoch": 5.81, "learning_rate": 2.5153030606121227e-06, "loss": 0.0043, "step": 18720 }, { "epoch": 5.81, "learning_rate": 2.5149029805961194e-06, "loss": 0.002, "step": 18721 }, { "epoch": 5.81, "learning_rate": 2.5145029005801165e-06, "loss": 0.0067, "step": 18722 }, { "epoch": 5.81, "learning_rate": 2.514102820564113e-06, "loss": 0.0042, "step": 18723 }, { "epoch": 5.81, "learning_rate": 2.5137027405481095e-06, "loss": 0.0039, "step": 18724 }, { "epoch": 5.81, "learning_rate": 2.5133026605321066e-06, "loss": 0.0149, "step": 18725 }, { "epoch": 5.81, "learning_rate": 2.5129025805161033e-06, "loss": 0.0168, "step": 18726 }, { "epoch": 5.81, "learning_rate": 2.5125025005001004e-06, "loss": 0.0107, "step": 18727 }, { "epoch": 5.81, "learning_rate": 2.512102420484097e-06, "loss": 0.0083, "step": 18728 }, { "epoch": 5.81, "learning_rate": 2.511702340468094e-06, "loss": 0.0311, "step": 18729 }, { "epoch": 5.81, "learning_rate": 2.5113022604520904e-06, "loss": 0.0435, "step": 18730 }, { "epoch": 5.81, "learning_rate": 2.510902180436087e-06, "loss": 0.0037, "step": 18731 }, { "epoch": 5.81, "learning_rate": 2.5105021004200842e-06, "loss": 0.0576, "step": 18732 }, { "epoch": 5.81, "learning_rate": 2.510102020404081e-06, "loss": 0.0035, "step": 18733 }, { "epoch": 5.81, "learning_rate": 2.509701940388078e-06, "loss": 0.0107, "step": 18734 }, { "epoch": 5.81, "learning_rate": 2.5093018603720747e-06, "loss": 0.0049, "step": 18735 }, { "epoch": 5.82, "learning_rate": 2.508901780356072e-06, "loss": 0.0037, "step": 18736 }, { "epoch": 5.82, "learning_rate": 2.508501700340068e-06, "loss": 0.0052, "step": 18737 }, { "epoch": 5.82, "learning_rate": 2.5081016203240648e-06, "loss": 0.0026, "step": 18738 }, { "epoch": 5.82, "learning_rate": 2.507701540308062e-06, "loss": 0.0096, "step": 18739 }, { "epoch": 5.82, "learning_rate": 2.5073014602920586e-06, "loss": 0.0074, "step": 18740 }, { "epoch": 5.82, "learning_rate": 2.5069013802760557e-06, "loss": 0.0043, "step": 18741 }, { "epoch": 5.82, "learning_rate": 2.5065013002600524e-06, "loss": 0.0105, "step": 18742 }, { "epoch": 5.82, "learning_rate": 2.5061012202440486e-06, "loss": 0.003, "step": 18743 }, { "epoch": 5.82, "learning_rate": 2.5057011402280457e-06, "loss": 0.007, "step": 18744 }, { "epoch": 5.82, "learning_rate": 2.5053010602120424e-06, "loss": 0.005, "step": 18745 }, { "epoch": 5.82, "learning_rate": 2.5049009801960395e-06, "loss": 0.009, "step": 18746 }, { "epoch": 5.82, "learning_rate": 2.5045009001800362e-06, "loss": 0.0117, "step": 18747 }, { "epoch": 5.82, "learning_rate": 2.5041008201640333e-06, "loss": 0.0081, "step": 18748 }, { "epoch": 5.82, "learning_rate": 2.5037007401480296e-06, "loss": 0.0086, "step": 18749 }, { "epoch": 5.82, "learning_rate": 2.5033006601320263e-06, "loss": 0.0081, "step": 18750 }, { "epoch": 5.82, "learning_rate": 2.5029005801160234e-06, "loss": 0.0373, "step": 18751 }, { "epoch": 5.82, "learning_rate": 2.50250050010002e-06, "loss": 0.0045, "step": 18752 }, { "epoch": 5.82, "learning_rate": 2.502100420084017e-06, "loss": 0.0036, "step": 18753 }, { "epoch": 5.82, "learning_rate": 2.501700340068014e-06, "loss": 0.0182, "step": 18754 }, { "epoch": 5.82, "learning_rate": 2.501300260052011e-06, "loss": 0.0219, "step": 18755 }, { "epoch": 5.82, "learning_rate": 2.5009001800360072e-06, "loss": 0.0075, "step": 18756 }, { "epoch": 5.82, "learning_rate": 2.500500100020004e-06, "loss": 0.0075, "step": 18757 }, { "epoch": 5.82, "learning_rate": 2.500100020004001e-06, "loss": 0.0044, "step": 18758 }, { "epoch": 5.82, "learning_rate": 2.4996999399879977e-06, "loss": 0.0032, "step": 18759 }, { "epoch": 5.82, "learning_rate": 2.4992998599719944e-06, "loss": 0.0025, "step": 18760 }, { "epoch": 5.82, "learning_rate": 2.4988997799559915e-06, "loss": 0.0101, "step": 18761 }, { "epoch": 5.82, "learning_rate": 2.498499699939988e-06, "loss": 0.0016, "step": 18762 }, { "epoch": 5.82, "learning_rate": 2.498099619923985e-06, "loss": 0.0171, "step": 18763 }, { "epoch": 5.82, "learning_rate": 2.4976995399079816e-06, "loss": 0.004, "step": 18764 }, { "epoch": 5.82, "learning_rate": 2.4972994598919787e-06, "loss": 0.0048, "step": 18765 }, { "epoch": 5.82, "learning_rate": 2.4968993798759754e-06, "loss": 0.0018, "step": 18766 }, { "epoch": 5.82, "learning_rate": 2.496499299859972e-06, "loss": 0.0083, "step": 18767 }, { "epoch": 5.83, "learning_rate": 2.496099219843969e-06, "loss": 0.0038, "step": 18768 }, { "epoch": 5.83, "learning_rate": 2.495699139827966e-06, "loss": 0.0187, "step": 18769 }, { "epoch": 5.83, "learning_rate": 2.4952990598119625e-06, "loss": 0.0035, "step": 18770 }, { "epoch": 5.83, "learning_rate": 2.4948989797959592e-06, "loss": 0.0844, "step": 18771 }, { "epoch": 5.83, "learning_rate": 2.4944988997799563e-06, "loss": 0.0039, "step": 18772 }, { "epoch": 5.83, "learning_rate": 2.494098819763953e-06, "loss": 0.0037, "step": 18773 }, { "epoch": 5.83, "learning_rate": 2.4936987397479497e-06, "loss": 0.0049, "step": 18774 }, { "epoch": 5.83, "learning_rate": 2.493298659731947e-06, "loss": 0.0101, "step": 18775 }, { "epoch": 5.83, "learning_rate": 2.492898579715943e-06, "loss": 0.0382, "step": 18776 }, { "epoch": 5.83, "learning_rate": 2.49249849969994e-06, "loss": 0.0045, "step": 18777 }, { "epoch": 5.83, "learning_rate": 2.492098419683937e-06, "loss": 0.0055, "step": 18778 }, { "epoch": 5.83, "learning_rate": 2.4916983396679336e-06, "loss": 0.005, "step": 18779 }, { "epoch": 5.83, "learning_rate": 2.4912982596519307e-06, "loss": 0.0158, "step": 18780 }, { "epoch": 5.83, "learning_rate": 2.4908981796359274e-06, "loss": 0.008, "step": 18781 }, { "epoch": 5.83, "learning_rate": 2.490498099619924e-06, "loss": 0.0033, "step": 18782 }, { "epoch": 5.83, "learning_rate": 2.490098019603921e-06, "loss": 0.0049, "step": 18783 }, { "epoch": 5.83, "learning_rate": 2.489697939587918e-06, "loss": 0.0051, "step": 18784 }, { "epoch": 5.83, "learning_rate": 2.4892978595719145e-06, "loss": 0.0062, "step": 18785 }, { "epoch": 5.83, "learning_rate": 2.4888977795559112e-06, "loss": 0.0062, "step": 18786 }, { "epoch": 5.83, "learning_rate": 2.4884976995399083e-06, "loss": 0.0172, "step": 18787 }, { "epoch": 5.83, "learning_rate": 2.488097619523905e-06, "loss": 0.0142, "step": 18788 }, { "epoch": 5.83, "learning_rate": 2.4876975395079017e-06, "loss": 0.0085, "step": 18789 }, { "epoch": 5.83, "learning_rate": 2.487297459491899e-06, "loss": 0.0045, "step": 18790 }, { "epoch": 5.83, "learning_rate": 2.4868973794758955e-06, "loss": 0.016, "step": 18791 }, { "epoch": 5.83, "learning_rate": 2.486497299459892e-06, "loss": 0.0055, "step": 18792 }, { "epoch": 5.83, "learning_rate": 2.486097219443889e-06, "loss": 0.0057, "step": 18793 }, { "epoch": 5.83, "learning_rate": 2.485697139427886e-06, "loss": 0.0064, "step": 18794 }, { "epoch": 5.83, "learning_rate": 2.4852970594118827e-06, "loss": 0.0082, "step": 18795 }, { "epoch": 5.83, "learning_rate": 2.4848969793958793e-06, "loss": 0.0013, "step": 18796 }, { "epoch": 5.83, "learning_rate": 2.4844968993798765e-06, "loss": 0.0059, "step": 18797 }, { "epoch": 5.83, "learning_rate": 2.4840968193638727e-06, "loss": 0.0028, "step": 18798 }, { "epoch": 5.83, "learning_rate": 2.48369673934787e-06, "loss": 0.007, "step": 18799 }, { "epoch": 5.84, "learning_rate": 2.4832966593318665e-06, "loss": 0.0096, "step": 18800 }, { "epoch": 5.84, "learning_rate": 2.482896579315863e-06, "loss": 0.0163, "step": 18801 }, { "epoch": 5.84, "learning_rate": 2.4824964992998603e-06, "loss": 0.0026, "step": 18802 }, { "epoch": 5.84, "learning_rate": 2.482096419283857e-06, "loss": 0.0018, "step": 18803 }, { "epoch": 5.84, "learning_rate": 2.4816963392678537e-06, "loss": 0.0105, "step": 18804 }, { "epoch": 5.84, "learning_rate": 2.4812962592518504e-06, "loss": 0.002, "step": 18805 }, { "epoch": 5.84, "learning_rate": 2.4808961792358475e-06, "loss": 0.0078, "step": 18806 }, { "epoch": 5.84, "learning_rate": 2.480496099219844e-06, "loss": 0.0188, "step": 18807 }, { "epoch": 5.84, "learning_rate": 2.480096019203841e-06, "loss": 0.0076, "step": 18808 }, { "epoch": 5.84, "learning_rate": 2.479695939187838e-06, "loss": 0.0023, "step": 18809 }, { "epoch": 5.84, "learning_rate": 2.4792958591718346e-06, "loss": 0.0072, "step": 18810 }, { "epoch": 5.84, "learning_rate": 2.4788957791558313e-06, "loss": 0.0026, "step": 18811 }, { "epoch": 5.84, "learning_rate": 2.478495699139828e-06, "loss": 0.0029, "step": 18812 }, { "epoch": 5.84, "learning_rate": 2.478095619123825e-06, "loss": 0.0032, "step": 18813 }, { "epoch": 5.84, "learning_rate": 2.477695539107822e-06, "loss": 0.0041, "step": 18814 }, { "epoch": 5.84, "learning_rate": 2.4772954590918185e-06, "loss": 0.0062, "step": 18815 }, { "epoch": 5.84, "learning_rate": 2.4768953790758156e-06, "loss": 0.0039, "step": 18816 }, { "epoch": 5.84, "learning_rate": 2.476495299059812e-06, "loss": 0.0036, "step": 18817 }, { "epoch": 5.84, "learning_rate": 2.476095219043809e-06, "loss": 0.002, "step": 18818 }, { "epoch": 5.84, "learning_rate": 2.4756951390278057e-06, "loss": 0.0059, "step": 18819 }, { "epoch": 5.84, "learning_rate": 2.4752950590118024e-06, "loss": 0.0032, "step": 18820 }, { "epoch": 5.84, "learning_rate": 2.4748949789957995e-06, "loss": 0.0026, "step": 18821 }, { "epoch": 5.84, "learning_rate": 2.474494898979796e-06, "loss": 0.0049, "step": 18822 }, { "epoch": 5.84, "learning_rate": 2.474094818963793e-06, "loss": 0.01, "step": 18823 }, { "epoch": 5.84, "learning_rate": 2.474094818963793e-06, "loss": 0.0383, "step": 18824 }, { "epoch": 5.84, "learning_rate": 2.4736947389477895e-06, "loss": 0.0076, "step": 18825 }, { "epoch": 5.84, "learning_rate": 2.4732946589317866e-06, "loss": 0.0257, "step": 18826 }, { "epoch": 5.84, "learning_rate": 2.4728945789157833e-06, "loss": 0.0037, "step": 18827 }, { "epoch": 5.84, "learning_rate": 2.47249449889978e-06, "loss": 0.0052, "step": 18828 }, { "epoch": 5.84, "learning_rate": 2.472094418883777e-06, "loss": 0.0221, "step": 18829 }, { "epoch": 5.84, "learning_rate": 2.471694338867774e-06, "loss": 0.0021, "step": 18830 }, { "epoch": 5.84, "learning_rate": 2.4712942588517705e-06, "loss": 0.0215, "step": 18831 }, { "epoch": 5.85, "learning_rate": 2.470894178835767e-06, "loss": 0.0126, "step": 18832 }, { "epoch": 5.85, "learning_rate": 2.4704940988197643e-06, "loss": 0.014, "step": 18833 }, { "epoch": 5.85, "learning_rate": 2.470094018803761e-06, "loss": 0.0042, "step": 18834 }, { "epoch": 5.85, "learning_rate": 2.4696939387877577e-06, "loss": 0.0168, "step": 18835 }, { "epoch": 5.85, "learning_rate": 2.4692938587717548e-06, "loss": 0.0031, "step": 18836 }, { "epoch": 5.85, "learning_rate": 2.4688937787557515e-06, "loss": 0.0172, "step": 18837 }, { "epoch": 5.85, "learning_rate": 2.468493698739748e-06, "loss": 0.0018, "step": 18838 }, { "epoch": 5.85, "learning_rate": 2.468093618723745e-06, "loss": 0.0075, "step": 18839 }, { "epoch": 5.85, "learning_rate": 2.4676935387077415e-06, "loss": 0.0183, "step": 18840 }, { "epoch": 5.85, "learning_rate": 2.4672934586917386e-06, "loss": 0.0014, "step": 18841 }, { "epoch": 5.85, "learning_rate": 2.4668933786757353e-06, "loss": 0.0043, "step": 18842 }, { "epoch": 5.85, "learning_rate": 2.466493298659732e-06, "loss": 0.0129, "step": 18843 }, { "epoch": 5.85, "learning_rate": 2.466093218643729e-06, "loss": 0.0043, "step": 18844 }, { "epoch": 5.85, "learning_rate": 2.465693138627726e-06, "loss": 0.0016, "step": 18845 }, { "epoch": 5.85, "learning_rate": 2.4652930586117225e-06, "loss": 0.0266, "step": 18846 }, { "epoch": 5.85, "learning_rate": 2.464892978595719e-06, "loss": 0.0018, "step": 18847 }, { "epoch": 5.85, "learning_rate": 2.4644928985797163e-06, "loss": 0.0051, "step": 18848 }, { "epoch": 5.85, "learning_rate": 2.464092818563713e-06, "loss": 0.0047, "step": 18849 }, { "epoch": 5.85, "learning_rate": 2.4636927385477096e-06, "loss": 0.0093, "step": 18850 }, { "epoch": 5.85, "learning_rate": 2.4632926585317068e-06, "loss": 0.0014, "step": 18851 }, { "epoch": 5.85, "learning_rate": 2.4628925785157034e-06, "loss": 0.0017, "step": 18852 }, { "epoch": 5.85, "learning_rate": 2.4624924984997e-06, "loss": 0.0063, "step": 18853 }, { "epoch": 5.85, "learning_rate": 2.462092418483697e-06, "loss": 0.0019, "step": 18854 }, { "epoch": 5.85, "learning_rate": 2.461692338467694e-06, "loss": 0.0048, "step": 18855 }, { "epoch": 5.85, "learning_rate": 2.4612922584516906e-06, "loss": 0.006, "step": 18856 }, { "epoch": 5.85, "learning_rate": 2.4608921784356873e-06, "loss": 0.0049, "step": 18857 }, { "epoch": 5.85, "learning_rate": 2.4604920984196844e-06, "loss": 0.0217, "step": 18858 }, { "epoch": 5.85, "learning_rate": 2.4600920184036807e-06, "loss": 0.0018, "step": 18859 }, { "epoch": 5.85, "learning_rate": 2.4596919383876778e-06, "loss": 0.037, "step": 18860 }, { "epoch": 5.85, "learning_rate": 2.4592918583716745e-06, "loss": 0.0438, "step": 18861 }, { "epoch": 5.85, "learning_rate": 2.458891778355671e-06, "loss": 0.0032, "step": 18862 }, { "epoch": 5.85, "learning_rate": 2.4584916983396683e-06, "loss": 0.0018, "step": 18863 }, { "epoch": 5.85, "learning_rate": 2.458091618323665e-06, "loss": 0.0032, "step": 18864 }, { "epoch": 5.86, "learning_rate": 2.4576915383076616e-06, "loss": 0.0063, "step": 18865 }, { "epoch": 5.86, "learning_rate": 2.4572914582916583e-06, "loss": 0.0064, "step": 18866 }, { "epoch": 5.86, "learning_rate": 2.4568913782756554e-06, "loss": 0.0183, "step": 18867 }, { "epoch": 5.86, "learning_rate": 2.456491298259652e-06, "loss": 0.0051, "step": 18868 }, { "epoch": 5.86, "learning_rate": 2.456091218243649e-06, "loss": 0.0035, "step": 18869 }, { "epoch": 5.86, "learning_rate": 2.455691138227646e-06, "loss": 0.0019, "step": 18870 }, { "epoch": 5.86, "learning_rate": 2.4552910582116426e-06, "loss": 0.0026, "step": 18871 }, { "epoch": 5.86, "learning_rate": 2.4548909781956393e-06, "loss": 0.0022, "step": 18872 }, { "epoch": 5.86, "learning_rate": 2.454490898179636e-06, "loss": 0.0016, "step": 18873 }, { "epoch": 5.86, "learning_rate": 2.454090818163633e-06, "loss": 0.0202, "step": 18874 }, { "epoch": 5.86, "learning_rate": 2.4536907381476298e-06, "loss": 0.0183, "step": 18875 }, { "epoch": 5.86, "learning_rate": 2.4532906581316265e-06, "loss": 0.0049, "step": 18876 }, { "epoch": 5.86, "learning_rate": 2.4528905781156236e-06, "loss": 0.015, "step": 18877 }, { "epoch": 5.86, "learning_rate": 2.45249049809962e-06, "loss": 0.0045, "step": 18878 }, { "epoch": 5.86, "learning_rate": 2.452090418083617e-06, "loss": 0.0044, "step": 18879 }, { "epoch": 5.86, "learning_rate": 2.4516903380676136e-06, "loss": 0.003, "step": 18880 }, { "epoch": 5.86, "learning_rate": 2.4512902580516103e-06, "loss": 0.0034, "step": 18881 }, { "epoch": 5.86, "learning_rate": 2.4508901780356074e-06, "loss": 0.0427, "step": 18882 }, { "epoch": 5.86, "learning_rate": 2.450490098019604e-06, "loss": 0.0037, "step": 18883 }, { "epoch": 5.86, "learning_rate": 2.450090018003601e-06, "loss": 0.0079, "step": 18884 }, { "epoch": 5.86, "learning_rate": 2.4496899379875975e-06, "loss": 0.0095, "step": 18885 }, { "epoch": 5.86, "learning_rate": 2.4492898579715946e-06, "loss": 0.0031, "step": 18886 }, { "epoch": 5.86, "learning_rate": 2.4488897779555913e-06, "loss": 0.0039, "step": 18887 }, { "epoch": 5.86, "learning_rate": 2.448489697939588e-06, "loss": 0.0151, "step": 18888 }, { "epoch": 5.86, "learning_rate": 2.448089617923585e-06, "loss": 0.0031, "step": 18889 }, { "epoch": 5.86, "learning_rate": 2.4476895379075818e-06, "loss": 0.0481, "step": 18890 }, { "epoch": 5.86, "learning_rate": 2.4472894578915784e-06, "loss": 0.0124, "step": 18891 }, { "epoch": 5.86, "learning_rate": 2.4468893778755756e-06, "loss": 0.0019, "step": 18892 }, { "epoch": 5.86, "learning_rate": 2.4464892978595722e-06, "loss": 0.0064, "step": 18893 }, { "epoch": 5.86, "learning_rate": 2.446089217843569e-06, "loss": 0.0024, "step": 18894 }, { "epoch": 5.86, "learning_rate": 2.4456891378275656e-06, "loss": 0.0237, "step": 18895 }, { "epoch": 5.86, "learning_rate": 2.4452890578115627e-06, "loss": 0.0071, "step": 18896 }, { "epoch": 5.87, "learning_rate": 2.4448889777955594e-06, "loss": 0.0111, "step": 18897 }, { "epoch": 5.87, "learning_rate": 2.444488897779556e-06, "loss": 0.0279, "step": 18898 }, { "epoch": 5.87, "learning_rate": 2.444088817763553e-06, "loss": 0.0041, "step": 18899 }, { "epoch": 5.87, "learning_rate": 2.4436887377475495e-06, "loss": 0.0097, "step": 18900 }, { "epoch": 5.87, "learning_rate": 2.4432886577315466e-06, "loss": 0.0048, "step": 18901 }, { "epoch": 5.87, "learning_rate": 2.4428885777155433e-06, "loss": 0.0149, "step": 18902 }, { "epoch": 5.87, "learning_rate": 2.44248849769954e-06, "loss": 0.0041, "step": 18903 }, { "epoch": 5.87, "learning_rate": 2.442088417683537e-06, "loss": 0.0061, "step": 18904 }, { "epoch": 5.87, "learning_rate": 2.4416883376675337e-06, "loss": 0.0183, "step": 18905 }, { "epoch": 5.87, "learning_rate": 2.4412882576515304e-06, "loss": 0.0106, "step": 18906 }, { "epoch": 5.87, "learning_rate": 2.440888177635527e-06, "loss": 0.0037, "step": 18907 }, { "epoch": 5.87, "learning_rate": 2.4404880976195242e-06, "loss": 0.0168, "step": 18908 }, { "epoch": 5.87, "learning_rate": 2.440088017603521e-06, "loss": 0.006, "step": 18909 }, { "epoch": 5.87, "learning_rate": 2.4396879375875176e-06, "loss": 0.0099, "step": 18910 }, { "epoch": 5.87, "learning_rate": 2.4392878575715147e-06, "loss": 0.007, "step": 18911 }, { "epoch": 5.87, "learning_rate": 2.4388877775555114e-06, "loss": 0.0052, "step": 18912 }, { "epoch": 5.87, "learning_rate": 2.438487697539508e-06, "loss": 0.002, "step": 18913 }, { "epoch": 5.87, "learning_rate": 2.4380876175235048e-06, "loss": 0.0031, "step": 18914 }, { "epoch": 5.87, "learning_rate": 2.437687537507502e-06, "loss": 0.0057, "step": 18915 }, { "epoch": 5.87, "learning_rate": 2.4372874574914986e-06, "loss": 0.0138, "step": 18916 }, { "epoch": 5.87, "learning_rate": 2.4368873774754953e-06, "loss": 0.0032, "step": 18917 }, { "epoch": 5.87, "learning_rate": 2.4364872974594924e-06, "loss": 0.0025, "step": 18918 }, { "epoch": 5.87, "learning_rate": 2.4360872174434886e-06, "loss": 0.0067, "step": 18919 }, { "epoch": 5.87, "learning_rate": 2.4356871374274857e-06, "loss": 0.0057, "step": 18920 }, { "epoch": 5.87, "learning_rate": 2.4352870574114824e-06, "loss": 0.0361, "step": 18921 }, { "epoch": 5.87, "learning_rate": 2.434886977395479e-06, "loss": 0.0015, "step": 18922 }, { "epoch": 5.87, "learning_rate": 2.4344868973794762e-06, "loss": 0.0028, "step": 18923 }, { "epoch": 5.87, "learning_rate": 2.434086817363473e-06, "loss": 0.0151, "step": 18924 }, { "epoch": 5.87, "learning_rate": 2.4336867373474696e-06, "loss": 0.0063, "step": 18925 }, { "epoch": 5.87, "learning_rate": 2.4332866573314663e-06, "loss": 0.0124, "step": 18926 }, { "epoch": 5.87, "learning_rate": 2.4328865773154634e-06, "loss": 0.013, "step": 18927 }, { "epoch": 5.87, "learning_rate": 2.43248649729946e-06, "loss": 0.0037, "step": 18928 }, { "epoch": 5.88, "learning_rate": 2.4320864172834568e-06, "loss": 0.0073, "step": 18929 }, { "epoch": 5.88, "learning_rate": 2.431686337267454e-06, "loss": 0.0096, "step": 18930 }, { "epoch": 5.88, "learning_rate": 2.4312862572514506e-06, "loss": 0.0046, "step": 18931 }, { "epoch": 5.88, "learning_rate": 2.4308861772354472e-06, "loss": 0.001, "step": 18932 }, { "epoch": 5.88, "learning_rate": 2.430486097219444e-06, "loss": 0.0082, "step": 18933 }, { "epoch": 5.88, "learning_rate": 2.430086017203441e-06, "loss": 0.0038, "step": 18934 }, { "epoch": 5.88, "learning_rate": 2.4296859371874377e-06, "loss": 0.0035, "step": 18935 }, { "epoch": 5.88, "learning_rate": 2.4292858571714344e-06, "loss": 0.0235, "step": 18936 }, { "epoch": 5.88, "learning_rate": 2.4288857771554315e-06, "loss": 0.0046, "step": 18937 }, { "epoch": 5.88, "learning_rate": 2.428485697139428e-06, "loss": 0.0058, "step": 18938 }, { "epoch": 5.88, "learning_rate": 2.428085617123425e-06, "loss": 0.006, "step": 18939 }, { "epoch": 5.88, "learning_rate": 2.4276855371074216e-06, "loss": 0.0052, "step": 18940 }, { "epoch": 5.88, "learning_rate": 2.4272854570914183e-06, "loss": 0.0014, "step": 18941 }, { "epoch": 5.88, "learning_rate": 2.4268853770754154e-06, "loss": 0.0029, "step": 18942 }, { "epoch": 5.88, "learning_rate": 2.426485297059412e-06, "loss": 0.0073, "step": 18943 }, { "epoch": 5.88, "learning_rate": 2.4260852170434087e-06, "loss": 0.0129, "step": 18944 }, { "epoch": 5.88, "learning_rate": 2.425685137027406e-06, "loss": 0.0071, "step": 18945 }, { "epoch": 5.88, "learning_rate": 2.4252850570114025e-06, "loss": 0.007, "step": 18946 }, { "epoch": 5.88, "learning_rate": 2.4248849769953992e-06, "loss": 0.0129, "step": 18947 }, { "epoch": 5.88, "learning_rate": 2.424484896979396e-06, "loss": 0.0094, "step": 18948 }, { "epoch": 5.88, "learning_rate": 2.424084816963393e-06, "loss": 0.0126, "step": 18949 }, { "epoch": 5.88, "learning_rate": 2.4236847369473897e-06, "loss": 0.0072, "step": 18950 }, { "epoch": 5.88, "learning_rate": 2.4232846569313864e-06, "loss": 0.0047, "step": 18951 }, { "epoch": 5.88, "learning_rate": 2.4228845769153835e-06, "loss": 0.0042, "step": 18952 }, { "epoch": 5.88, "learning_rate": 2.42248449689938e-06, "loss": 0.0023, "step": 18953 }, { "epoch": 5.88, "learning_rate": 2.422084416883377e-06, "loss": 0.0059, "step": 18954 }, { "epoch": 5.88, "learning_rate": 2.4216843368673736e-06, "loss": 0.0038, "step": 18955 }, { "epoch": 5.88, "learning_rate": 2.4212842568513707e-06, "loss": 0.004, "step": 18956 }, { "epoch": 5.88, "learning_rate": 2.4208841768353674e-06, "loss": 0.0043, "step": 18957 }, { "epoch": 5.88, "learning_rate": 2.420484096819364e-06, "loss": 0.0049, "step": 18958 }, { "epoch": 5.88, "learning_rate": 2.420084016803361e-06, "loss": 0.0159, "step": 18959 }, { "epoch": 5.88, "learning_rate": 2.4196839367873574e-06, "loss": 0.0013, "step": 18960 }, { "epoch": 5.89, "learning_rate": 2.4192838567713545e-06, "loss": 0.0353, "step": 18961 }, { "epoch": 5.89, "learning_rate": 2.4188837767553512e-06, "loss": 0.0072, "step": 18962 }, { "epoch": 5.89, "learning_rate": 2.418483696739348e-06, "loss": 0.0058, "step": 18963 }, { "epoch": 5.89, "learning_rate": 2.418083616723345e-06, "loss": 0.0053, "step": 18964 }, { "epoch": 5.89, "learning_rate": 2.4176835367073417e-06, "loss": 0.0037, "step": 18965 }, { "epoch": 5.89, "learning_rate": 2.4172834566913384e-06, "loss": 0.0038, "step": 18966 }, { "epoch": 5.89, "learning_rate": 2.416883376675335e-06, "loss": 0.0041, "step": 18967 }, { "epoch": 5.89, "learning_rate": 2.416483296659332e-06, "loss": 0.0113, "step": 18968 }, { "epoch": 5.89, "learning_rate": 2.416083216643329e-06, "loss": 0.006, "step": 18969 }, { "epoch": 5.89, "learning_rate": 2.4156831366273256e-06, "loss": 0.0052, "step": 18970 }, { "epoch": 5.89, "learning_rate": 2.4152830566113227e-06, "loss": 0.0058, "step": 18971 }, { "epoch": 5.89, "learning_rate": 2.4148829765953194e-06, "loss": 0.0082, "step": 18972 }, { "epoch": 5.89, "learning_rate": 2.414482896579316e-06, "loss": 0.0073, "step": 18973 }, { "epoch": 5.89, "learning_rate": 2.4140828165633127e-06, "loss": 0.0097, "step": 18974 }, { "epoch": 5.89, "learning_rate": 2.41368273654731e-06, "loss": 0.0027, "step": 18975 }, { "epoch": 5.89, "learning_rate": 2.4132826565313065e-06, "loss": 0.0026, "step": 18976 }, { "epoch": 5.89, "learning_rate": 2.412882576515303e-06, "loss": 0.0324, "step": 18977 }, { "epoch": 5.89, "learning_rate": 2.4124824964993003e-06, "loss": 0.0061, "step": 18978 }, { "epoch": 5.89, "learning_rate": 2.4120824164832966e-06, "loss": 0.0036, "step": 18979 }, { "epoch": 5.89, "learning_rate": 2.4116823364672937e-06, "loss": 0.0048, "step": 18980 }, { "epoch": 5.89, "learning_rate": 2.4112822564512904e-06, "loss": 0.0016, "step": 18981 }, { "epoch": 5.89, "learning_rate": 2.410882176435287e-06, "loss": 0.007, "step": 18982 }, { "epoch": 5.89, "learning_rate": 2.410482096419284e-06, "loss": 0.005, "step": 18983 }, { "epoch": 5.89, "learning_rate": 2.410082016403281e-06, "loss": 0.0027, "step": 18984 }, { "epoch": 5.89, "learning_rate": 2.4096819363872775e-06, "loss": 0.003, "step": 18985 }, { "epoch": 5.89, "learning_rate": 2.4092818563712742e-06, "loss": 0.0123, "step": 18986 }, { "epoch": 5.89, "learning_rate": 2.4088817763552713e-06, "loss": 0.0043, "step": 18987 }, { "epoch": 5.89, "learning_rate": 2.408481696339268e-06, "loss": 0.0022, "step": 18988 }, { "epoch": 5.89, "learning_rate": 2.4080816163232647e-06, "loss": 0.0087, "step": 18989 }, { "epoch": 5.89, "learning_rate": 2.407681536307262e-06, "loss": 0.0062, "step": 18990 }, { "epoch": 5.89, "learning_rate": 2.4072814562912585e-06, "loss": 0.0137, "step": 18991 }, { "epoch": 5.89, "learning_rate": 2.406881376275255e-06, "loss": 0.0018, "step": 18992 }, { "epoch": 5.9, "learning_rate": 2.406481296259252e-06, "loss": 0.01, "step": 18993 }, { "epoch": 5.9, "learning_rate": 2.406081216243249e-06, "loss": 0.0038, "step": 18994 }, { "epoch": 5.9, "learning_rate": 2.4056811362272457e-06, "loss": 0.0188, "step": 18995 }, { "epoch": 5.9, "learning_rate": 2.4052810562112424e-06, "loss": 0.0023, "step": 18996 }, { "epoch": 5.9, "learning_rate": 2.4048809761952395e-06, "loss": 0.0025, "step": 18997 }, { "epoch": 5.9, "learning_rate": 2.404480896179236e-06, "loss": 0.0018, "step": 18998 }, { "epoch": 5.9, "learning_rate": 2.404080816163233e-06, "loss": 0.0028, "step": 18999 }, { "epoch": 5.9, "learning_rate": 2.40368073614723e-06, "loss": 0.0104, "step": 19000 }, { "epoch": 5.9, "learning_rate": 2.4032806561312262e-06, "loss": 0.0184, "step": 19001 }, { "epoch": 5.9, "learning_rate": 2.4028805761152233e-06, "loss": 0.0123, "step": 19002 }, { "epoch": 5.9, "learning_rate": 2.40248049609922e-06, "loss": 0.0009, "step": 19003 }, { "epoch": 5.9, "learning_rate": 2.4020804160832167e-06, "loss": 0.0015, "step": 19004 }, { "epoch": 5.9, "learning_rate": 2.401680336067214e-06, "loss": 0.0021, "step": 19005 }, { "epoch": 5.9, "learning_rate": 2.4012802560512105e-06, "loss": 0.0478, "step": 19006 }, { "epoch": 5.9, "learning_rate": 2.400880176035207e-06, "loss": 0.0093, "step": 19007 }, { "epoch": 5.9, "learning_rate": 2.400480096019204e-06, "loss": 0.0028, "step": 19008 }, { "epoch": 5.9, "learning_rate": 2.400080016003201e-06, "loss": 0.0117, "step": 19009 }, { "epoch": 5.9, "learning_rate": 2.3996799359871977e-06, "loss": 0.0054, "step": 19010 }, { "epoch": 5.9, "learning_rate": 2.3992798559711944e-06, "loss": 0.0074, "step": 19011 }, { "epoch": 5.9, "learning_rate": 2.3988797759551915e-06, "loss": 0.0487, "step": 19012 }, { "epoch": 5.9, "learning_rate": 2.398479695939188e-06, "loss": 0.0087, "step": 19013 }, { "epoch": 5.9, "learning_rate": 2.398079615923185e-06, "loss": 0.0184, "step": 19014 }, { "epoch": 5.9, "learning_rate": 2.3976795359071815e-06, "loss": 0.0059, "step": 19015 }, { "epoch": 5.9, "learning_rate": 2.3972794558911786e-06, "loss": 0.0077, "step": 19016 }, { "epoch": 5.9, "learning_rate": 2.3968793758751753e-06, "loss": 0.0156, "step": 19017 }, { "epoch": 5.9, "learning_rate": 2.396479295859172e-06, "loss": 0.004, "step": 19018 }, { "epoch": 5.9, "learning_rate": 2.396079215843169e-06, "loss": 0.0086, "step": 19019 }, { "epoch": 5.9, "learning_rate": 2.3956791358271654e-06, "loss": 0.0027, "step": 19020 }, { "epoch": 5.9, "learning_rate": 2.3952790558111625e-06, "loss": 0.0035, "step": 19021 }, { "epoch": 5.9, "learning_rate": 2.394878975795159e-06, "loss": 0.0202, "step": 19022 }, { "epoch": 5.9, "learning_rate": 2.394478895779156e-06, "loss": 0.0093, "step": 19023 }, { "epoch": 5.9, "learning_rate": 2.394078815763153e-06, "loss": 0.0051, "step": 19024 }, { "epoch": 5.9, "learning_rate": 2.3936787357471497e-06, "loss": 0.0046, "step": 19025 }, { "epoch": 5.91, "learning_rate": 2.3932786557311463e-06, "loss": 0.0029, "step": 19026 }, { "epoch": 5.91, "learning_rate": 2.392878575715143e-06, "loss": 0.0026, "step": 19027 }, { "epoch": 5.91, "learning_rate": 2.39247849569914e-06, "loss": 0.004, "step": 19028 }, { "epoch": 5.91, "learning_rate": 2.392078415683137e-06, "loss": 0.0364, "step": 19029 }, { "epoch": 5.91, "learning_rate": 2.3916783356671335e-06, "loss": 0.0129, "step": 19030 }, { "epoch": 5.91, "learning_rate": 2.3912782556511306e-06, "loss": 0.0079, "step": 19031 }, { "epoch": 5.91, "learning_rate": 2.390878175635127e-06, "loss": 0.0171, "step": 19032 }, { "epoch": 5.91, "learning_rate": 2.390478095619124e-06, "loss": 0.0141, "step": 19033 }, { "epoch": 5.91, "learning_rate": 2.3900780156031207e-06, "loss": 0.002, "step": 19034 }, { "epoch": 5.91, "learning_rate": 2.3896779355871178e-06, "loss": 0.0022, "step": 19035 }, { "epoch": 5.91, "learning_rate": 2.3892778555711145e-06, "loss": 0.0036, "step": 19036 }, { "epoch": 5.91, "learning_rate": 2.388877775555111e-06, "loss": 0.0067, "step": 19037 }, { "epoch": 5.91, "learning_rate": 2.3884776955391083e-06, "loss": 0.0153, "step": 19038 }, { "epoch": 5.91, "learning_rate": 2.3880776155231045e-06, "loss": 0.0213, "step": 19039 }, { "epoch": 5.91, "learning_rate": 2.3876775355071016e-06, "loss": 0.0052, "step": 19040 }, { "epoch": 5.91, "learning_rate": 2.3872774554910983e-06, "loss": 0.0691, "step": 19041 }, { "epoch": 5.91, "learning_rate": 2.386877375475095e-06, "loss": 0.0063, "step": 19042 }, { "epoch": 5.91, "learning_rate": 2.386477295459092e-06, "loss": 0.0132, "step": 19043 }, { "epoch": 5.91, "learning_rate": 2.386077215443089e-06, "loss": 0.0362, "step": 19044 }, { "epoch": 5.91, "learning_rate": 2.3856771354270855e-06, "loss": 0.011, "step": 19045 }, { "epoch": 5.91, "learning_rate": 2.3852770554110826e-06, "loss": 0.0033, "step": 19046 }, { "epoch": 5.91, "learning_rate": 2.3848769753950793e-06, "loss": 0.0049, "step": 19047 }, { "epoch": 5.91, "learning_rate": 2.384476895379076e-06, "loss": 0.0157, "step": 19048 }, { "epoch": 5.91, "learning_rate": 2.3840768153630727e-06, "loss": 0.001, "step": 19049 }, { "epoch": 5.91, "learning_rate": 2.3836767353470698e-06, "loss": 0.0098, "step": 19050 }, { "epoch": 5.91, "learning_rate": 2.3832766553310665e-06, "loss": 0.0076, "step": 19051 }, { "epoch": 5.91, "learning_rate": 2.382876575315063e-06, "loss": 0.0137, "step": 19052 }, { "epoch": 5.91, "learning_rate": 2.3824764952990603e-06, "loss": 0.0049, "step": 19053 }, { "epoch": 5.91, "learning_rate": 2.382076415283057e-06, "loss": 0.0156, "step": 19054 }, { "epoch": 5.91, "learning_rate": 2.3816763352670536e-06, "loss": 0.0152, "step": 19055 }, { "epoch": 5.91, "learning_rate": 2.3812762552510503e-06, "loss": 0.0131, "step": 19056 }, { "epoch": 5.91, "learning_rate": 2.3808761752350474e-06, "loss": 0.0055, "step": 19057 }, { "epoch": 5.92, "learning_rate": 2.380476095219044e-06, "loss": 0.0023, "step": 19058 }, { "epoch": 5.92, "learning_rate": 2.380076015203041e-06, "loss": 0.004, "step": 19059 }, { "epoch": 5.92, "learning_rate": 2.379675935187038e-06, "loss": 0.003, "step": 19060 }, { "epoch": 5.92, "learning_rate": 2.379275855171034e-06, "loss": 0.0023, "step": 19061 }, { "epoch": 5.92, "learning_rate": 2.3788757751550313e-06, "loss": 0.0089, "step": 19062 }, { "epoch": 5.92, "learning_rate": 2.378475695139028e-06, "loss": 0.0037, "step": 19063 }, { "epoch": 5.92, "learning_rate": 2.3780756151230247e-06, "loss": 0.0076, "step": 19064 }, { "epoch": 5.92, "learning_rate": 2.3776755351070218e-06, "loss": 0.0052, "step": 19065 }, { "epoch": 5.92, "learning_rate": 2.3772754550910184e-06, "loss": 0.0061, "step": 19066 }, { "epoch": 5.92, "learning_rate": 2.376875375075015e-06, "loss": 0.0172, "step": 19067 }, { "epoch": 5.92, "learning_rate": 2.376475295059012e-06, "loss": 0.0021, "step": 19068 }, { "epoch": 5.92, "learning_rate": 2.376075215043009e-06, "loss": 0.0029, "step": 19069 }, { "epoch": 5.92, "learning_rate": 2.3756751350270056e-06, "loss": 0.0027, "step": 19070 }, { "epoch": 5.92, "learning_rate": 2.3752750550110023e-06, "loss": 0.0034, "step": 19071 }, { "epoch": 5.92, "learning_rate": 2.3748749749949994e-06, "loss": 0.0018, "step": 19072 }, { "epoch": 5.92, "learning_rate": 2.3744748949789957e-06, "loss": 0.003, "step": 19073 }, { "epoch": 5.92, "learning_rate": 2.3740748149629928e-06, "loss": 0.0025, "step": 19074 }, { "epoch": 5.92, "learning_rate": 2.3736747349469895e-06, "loss": 0.0136, "step": 19075 }, { "epoch": 5.92, "learning_rate": 2.3732746549309866e-06, "loss": 0.0017, "step": 19076 }, { "epoch": 5.92, "learning_rate": 2.3728745749149833e-06, "loss": 0.0109, "step": 19077 }, { "epoch": 5.92, "learning_rate": 2.37247449489898e-06, "loss": 0.0028, "step": 19078 }, { "epoch": 5.92, "learning_rate": 2.372074414882977e-06, "loss": 0.0088, "step": 19079 }, { "epoch": 5.92, "learning_rate": 2.3716743348669733e-06, "loss": 0.0125, "step": 19080 }, { "epoch": 5.92, "learning_rate": 2.3712742548509704e-06, "loss": 0.0391, "step": 19081 }, { "epoch": 5.92, "learning_rate": 2.370874174834967e-06, "loss": 0.0053, "step": 19082 }, { "epoch": 5.92, "learning_rate": 2.370474094818964e-06, "loss": 0.0038, "step": 19083 }, { "epoch": 5.92, "learning_rate": 2.370074014802961e-06, "loss": 0.0126, "step": 19084 }, { "epoch": 5.92, "learning_rate": 2.3696739347869576e-06, "loss": 0.0059, "step": 19085 }, { "epoch": 5.92, "learning_rate": 2.3692738547709543e-06, "loss": 0.0075, "step": 19086 }, { "epoch": 5.92, "learning_rate": 2.368873774754951e-06, "loss": 0.0047, "step": 19087 }, { "epoch": 5.92, "learning_rate": 2.368473694738948e-06, "loss": 0.0058, "step": 19088 }, { "epoch": 5.92, "learning_rate": 2.3680736147229448e-06, "loss": 0.023, "step": 19089 }, { "epoch": 5.93, "learning_rate": 2.3676735347069415e-06, "loss": 0.0052, "step": 19090 }, { "epoch": 5.93, "learning_rate": 2.3672734546909386e-06, "loss": 0.0186, "step": 19091 }, { "epoch": 5.93, "learning_rate": 2.3668733746749353e-06, "loss": 0.0787, "step": 19092 }, { "epoch": 5.93, "learning_rate": 2.366473294658932e-06, "loss": 0.0159, "step": 19093 }, { "epoch": 5.93, "learning_rate": 2.3660732146429286e-06, "loss": 0.004, "step": 19094 }, { "epoch": 5.93, "learning_rate": 2.3656731346269257e-06, "loss": 0.0227, "step": 19095 }, { "epoch": 5.93, "learning_rate": 2.3652730546109224e-06, "loss": 0.0036, "step": 19096 }, { "epoch": 5.93, "learning_rate": 2.364872974594919e-06, "loss": 0.0083, "step": 19097 }, { "epoch": 5.93, "learning_rate": 2.3644728945789162e-06, "loss": 0.0092, "step": 19098 }, { "epoch": 5.93, "learning_rate": 2.364072814562913e-06, "loss": 0.0037, "step": 19099 }, { "epoch": 5.93, "learning_rate": 2.3636727345469096e-06, "loss": 0.0413, "step": 19100 }, { "epoch": 5.93, "learning_rate": 2.3632726545309063e-06, "loss": 0.0022, "step": 19101 }, { "epoch": 5.93, "learning_rate": 2.362872574514903e-06, "loss": 0.0047, "step": 19102 }, { "epoch": 5.93, "learning_rate": 2.3624724944989e-06, "loss": 0.0024, "step": 19103 }, { "epoch": 5.93, "learning_rate": 2.3620724144828968e-06, "loss": 0.004, "step": 19104 }, { "epoch": 5.93, "learning_rate": 2.3616723344668934e-06, "loss": 0.013, "step": 19105 }, { "epoch": 5.93, "learning_rate": 2.3612722544508906e-06, "loss": 0.012, "step": 19106 }, { "epoch": 5.93, "learning_rate": 2.3608721744348872e-06, "loss": 0.0038, "step": 19107 }, { "epoch": 5.93, "learning_rate": 2.360472094418884e-06, "loss": 0.0024, "step": 19108 }, { "epoch": 5.93, "learning_rate": 2.3600720144028806e-06, "loss": 0.004, "step": 19109 }, { "epoch": 5.93, "learning_rate": 2.3596719343868777e-06, "loss": 0.0026, "step": 19110 }, { "epoch": 5.93, "learning_rate": 2.3592718543708744e-06, "loss": 0.0023, "step": 19111 }, { "epoch": 5.93, "learning_rate": 2.358871774354871e-06, "loss": 0.0746, "step": 19112 }, { "epoch": 5.93, "learning_rate": 2.358471694338868e-06, "loss": 0.0048, "step": 19113 }, { "epoch": 5.93, "learning_rate": 2.3580716143228645e-06, "loss": 0.0028, "step": 19114 }, { "epoch": 5.93, "learning_rate": 2.3576715343068616e-06, "loss": 0.0021, "step": 19115 }, { "epoch": 5.93, "learning_rate": 2.3572714542908583e-06, "loss": 0.0022, "step": 19116 }, { "epoch": 5.93, "learning_rate": 2.3568713742748554e-06, "loss": 0.0138, "step": 19117 }, { "epoch": 5.93, "learning_rate": 2.356471294258852e-06, "loss": 0.0017, "step": 19118 }, { "epoch": 5.93, "learning_rate": 2.3560712142428488e-06, "loss": 0.0099, "step": 19119 }, { "epoch": 5.93, "learning_rate": 2.355671134226846e-06, "loss": 0.0017, "step": 19120 }, { "epoch": 5.93, "learning_rate": 2.355271054210842e-06, "loss": 0.0064, "step": 19121 }, { "epoch": 5.94, "learning_rate": 2.3548709741948392e-06, "loss": 0.0082, "step": 19122 }, { "epoch": 5.94, "learning_rate": 2.354470894178836e-06, "loss": 0.0042, "step": 19123 }, { "epoch": 5.94, "learning_rate": 2.3540708141628326e-06, "loss": 0.003, "step": 19124 }, { "epoch": 5.94, "learning_rate": 2.3536707341468297e-06, "loss": 0.0036, "step": 19125 }, { "epoch": 5.94, "learning_rate": 2.3532706541308264e-06, "loss": 0.0087, "step": 19126 }, { "epoch": 5.94, "learning_rate": 2.352870574114823e-06, "loss": 0.0533, "step": 19127 }, { "epoch": 5.94, "learning_rate": 2.3524704940988198e-06, "loss": 0.0038, "step": 19128 }, { "epoch": 5.94, "learning_rate": 2.352070414082817e-06, "loss": 0.0061, "step": 19129 }, { "epoch": 5.94, "learning_rate": 2.3516703340668136e-06, "loss": 0.0029, "step": 19130 }, { "epoch": 5.94, "learning_rate": 2.3512702540508103e-06, "loss": 0.0072, "step": 19131 }, { "epoch": 5.94, "learning_rate": 2.3508701740348074e-06, "loss": 0.0066, "step": 19132 }, { "epoch": 5.94, "learning_rate": 2.3504700940188036e-06, "loss": 0.002, "step": 19133 }, { "epoch": 5.94, "learning_rate": 2.3500700140028007e-06, "loss": 0.0034, "step": 19134 }, { "epoch": 5.94, "learning_rate": 2.3496699339867974e-06, "loss": 0.0042, "step": 19135 }, { "epoch": 5.94, "learning_rate": 2.3492698539707945e-06, "loss": 0.0019, "step": 19136 }, { "epoch": 5.94, "learning_rate": 2.3488697739547912e-06, "loss": 0.0045, "step": 19137 }, { "epoch": 5.94, "learning_rate": 2.348469693938788e-06, "loss": 0.0046, "step": 19138 }, { "epoch": 5.94, "learning_rate": 2.348069613922785e-06, "loss": 0.0072, "step": 19139 }, { "epoch": 5.94, "learning_rate": 2.3476695339067813e-06, "loss": 0.0342, "step": 19140 }, { "epoch": 5.94, "learning_rate": 2.3472694538907784e-06, "loss": 0.0659, "step": 19141 }, { "epoch": 5.94, "learning_rate": 2.346869373874775e-06, "loss": 0.0018, "step": 19142 }, { "epoch": 5.94, "learning_rate": 2.3464692938587718e-06, "loss": 0.0643, "step": 19143 }, { "epoch": 5.94, "learning_rate": 2.346069213842769e-06, "loss": 0.0084, "step": 19144 }, { "epoch": 5.94, "learning_rate": 2.3456691338267656e-06, "loss": 0.003, "step": 19145 }, { "epoch": 5.94, "learning_rate": 2.3452690538107622e-06, "loss": 0.0126, "step": 19146 }, { "epoch": 5.94, "learning_rate": 2.344868973794759e-06, "loss": 0.0236, "step": 19147 }, { "epoch": 5.94, "learning_rate": 2.344468893778756e-06, "loss": 0.0787, "step": 19148 }, { "epoch": 5.94, "learning_rate": 2.3440688137627527e-06, "loss": 0.0077, "step": 19149 }, { "epoch": 5.94, "learning_rate": 2.3436687337467494e-06, "loss": 0.0016, "step": 19150 }, { "epoch": 5.94, "learning_rate": 2.3432686537307465e-06, "loss": 0.0269, "step": 19151 }, { "epoch": 5.94, "learning_rate": 2.342868573714743e-06, "loss": 0.0025, "step": 19152 }, { "epoch": 5.94, "learning_rate": 2.34246849369874e-06, "loss": 0.0053, "step": 19153 }, { "epoch": 5.94, "learning_rate": 2.342068413682737e-06, "loss": 0.015, "step": 19154 }, { "epoch": 5.95, "learning_rate": 2.3416683336667333e-06, "loss": 0.0099, "step": 19155 }, { "epoch": 5.95, "learning_rate": 2.3412682536507304e-06, "loss": 0.0055, "step": 19156 }, { "epoch": 5.95, "learning_rate": 2.340868173634727e-06, "loss": 0.012, "step": 19157 }, { "epoch": 5.95, "learning_rate": 2.340468093618724e-06, "loss": 0.0034, "step": 19158 }, { "epoch": 5.95, "learning_rate": 2.340068013602721e-06, "loss": 0.0157, "step": 19159 }, { "epoch": 5.95, "learning_rate": 2.3396679335867175e-06, "loss": 0.0045, "step": 19160 }, { "epoch": 5.95, "learning_rate": 2.3392678535707147e-06, "loss": 0.0045, "step": 19161 }, { "epoch": 5.95, "learning_rate": 2.338867773554711e-06, "loss": 0.0092, "step": 19162 }, { "epoch": 5.95, "learning_rate": 2.338467693538708e-06, "loss": 0.0027, "step": 19163 }, { "epoch": 5.95, "learning_rate": 2.3380676135227047e-06, "loss": 0.0088, "step": 19164 }, { "epoch": 5.95, "learning_rate": 2.3376675335067014e-06, "loss": 0.0038, "step": 19165 }, { "epoch": 5.95, "learning_rate": 2.3372674534906985e-06, "loss": 0.0034, "step": 19166 }, { "epoch": 5.95, "learning_rate": 2.336867373474695e-06, "loss": 0.0189, "step": 19167 }, { "epoch": 5.95, "learning_rate": 2.336467293458692e-06, "loss": 0.0043, "step": 19168 }, { "epoch": 5.95, "learning_rate": 2.3360672134426886e-06, "loss": 0.0086, "step": 19169 }, { "epoch": 5.95, "learning_rate": 2.3356671334266857e-06, "loss": 0.0043, "step": 19170 }, { "epoch": 5.95, "learning_rate": 2.3352670534106824e-06, "loss": 0.006, "step": 19171 }, { "epoch": 5.95, "learning_rate": 2.334866973394679e-06, "loss": 0.0057, "step": 19172 }, { "epoch": 5.95, "learning_rate": 2.334466893378676e-06, "loss": 0.0017, "step": 19173 }, { "epoch": 5.95, "learning_rate": 2.3340668133626724e-06, "loss": 0.0041, "step": 19174 }, { "epoch": 5.95, "learning_rate": 2.3336667333466695e-06, "loss": 0.0109, "step": 19175 }, { "epoch": 5.95, "learning_rate": 2.3332666533306662e-06, "loss": 0.0126, "step": 19176 }, { "epoch": 5.95, "learning_rate": 2.332866573314663e-06, "loss": 0.0054, "step": 19177 }, { "epoch": 5.95, "learning_rate": 2.33246649329866e-06, "loss": 0.0063, "step": 19178 }, { "epoch": 5.95, "learning_rate": 2.3320664132826567e-06, "loss": 0.0078, "step": 19179 }, { "epoch": 5.95, "learning_rate": 2.331666333266654e-06, "loss": 0.026, "step": 19180 }, { "epoch": 5.95, "learning_rate": 2.33126625325065e-06, "loss": 0.0661, "step": 19181 }, { "epoch": 5.95, "learning_rate": 2.330866173234647e-06, "loss": 0.0121, "step": 19182 }, { "epoch": 5.95, "learning_rate": 2.330466093218644e-06, "loss": 0.0315, "step": 19183 }, { "epoch": 5.95, "learning_rate": 2.3300660132026406e-06, "loss": 0.0141, "step": 19184 }, { "epoch": 5.95, "learning_rate": 2.3296659331866377e-06, "loss": 0.0113, "step": 19185 }, { "epoch": 5.95, "learning_rate": 2.3292658531706344e-06, "loss": 0.0055, "step": 19186 }, { "epoch": 5.96, "learning_rate": 2.328865773154631e-06, "loss": 0.0122, "step": 19187 }, { "epoch": 5.96, "learning_rate": 2.3284656931386277e-06, "loss": 0.0112, "step": 19188 }, { "epoch": 5.96, "learning_rate": 2.328065613122625e-06, "loss": 0.0028, "step": 19189 }, { "epoch": 5.96, "learning_rate": 2.3276655331066215e-06, "loss": 0.007, "step": 19190 }, { "epoch": 5.96, "learning_rate": 2.327265453090618e-06, "loss": 0.0026, "step": 19191 }, { "epoch": 5.96, "learning_rate": 2.3268653730746153e-06, "loss": 0.0074, "step": 19192 }, { "epoch": 5.96, "learning_rate": 2.326465293058612e-06, "loss": 0.0034, "step": 19193 }, { "epoch": 5.96, "learning_rate": 2.3260652130426087e-06, "loss": 0.0669, "step": 19194 }, { "epoch": 5.96, "learning_rate": 2.3256651330266054e-06, "loss": 0.0075, "step": 19195 }, { "epoch": 5.96, "learning_rate": 2.325265053010602e-06, "loss": 0.0035, "step": 19196 }, { "epoch": 5.96, "learning_rate": 2.324864972994599e-06, "loss": 0.014, "step": 19197 }, { "epoch": 5.96, "learning_rate": 2.324464892978596e-06, "loss": 0.034, "step": 19198 }, { "epoch": 5.96, "learning_rate": 2.324064812962593e-06, "loss": 0.0148, "step": 19199 }, { "epoch": 5.96, "learning_rate": 2.3236647329465897e-06, "loss": 0.0077, "step": 19200 }, { "epoch": 5.96, "learning_rate": 2.3232646529305863e-06, "loss": 0.0061, "step": 19201 }, { "epoch": 5.96, "learning_rate": 2.322864572914583e-06, "loss": 0.0969, "step": 19202 }, { "epoch": 5.96, "learning_rate": 2.3224644928985797e-06, "loss": 0.0137, "step": 19203 }, { "epoch": 5.96, "learning_rate": 2.322064412882577e-06, "loss": 0.0104, "step": 19204 }, { "epoch": 5.96, "learning_rate": 2.3216643328665735e-06, "loss": 0.0153, "step": 19205 }, { "epoch": 5.96, "learning_rate": 2.32126425285057e-06, "loss": 0.0036, "step": 19206 }, { "epoch": 5.96, "learning_rate": 2.3208641728345673e-06, "loss": 0.005, "step": 19207 }, { "epoch": 5.96, "learning_rate": 2.320464092818564e-06, "loss": 0.0051, "step": 19208 }, { "epoch": 5.96, "learning_rate": 2.3200640128025607e-06, "loss": 0.01, "step": 19209 }, { "epoch": 5.96, "learning_rate": 2.3196639327865574e-06, "loss": 0.004, "step": 19210 }, { "epoch": 5.96, "learning_rate": 2.3192638527705545e-06, "loss": 0.0067, "step": 19211 }, { "epoch": 5.96, "learning_rate": 2.318863772754551e-06, "loss": 0.0025, "step": 19212 }, { "epoch": 5.96, "learning_rate": 2.318463692738548e-06, "loss": 0.0038, "step": 19213 }, { "epoch": 5.96, "learning_rate": 2.318063612722545e-06, "loss": 0.0053, "step": 19214 }, { "epoch": 5.96, "learning_rate": 2.3176635327065412e-06, "loss": 0.0058, "step": 19215 }, { "epoch": 5.96, "learning_rate": 2.3172634526905383e-06, "loss": 0.0595, "step": 19216 }, { "epoch": 5.96, "learning_rate": 2.316863372674535e-06, "loss": 0.0029, "step": 19217 }, { "epoch": 5.96, "learning_rate": 2.3164632926585317e-06, "loss": 0.0074, "step": 19218 }, { "epoch": 5.97, "learning_rate": 2.316063212642529e-06, "loss": 0.0099, "step": 19219 }, { "epoch": 5.97, "learning_rate": 2.3156631326265255e-06, "loss": 0.004, "step": 19220 }, { "epoch": 5.97, "learning_rate": 2.3152630526105226e-06, "loss": 0.0132, "step": 19221 }, { "epoch": 5.97, "learning_rate": 2.314862972594519e-06, "loss": 0.014, "step": 19222 }, { "epoch": 5.97, "learning_rate": 2.314462892578516e-06, "loss": 0.0028, "step": 19223 }, { "epoch": 5.97, "learning_rate": 2.3140628125625127e-06, "loss": 0.045, "step": 19224 }, { "epoch": 5.97, "learning_rate": 2.3136627325465094e-06, "loss": 0.0063, "step": 19225 }, { "epoch": 5.97, "learning_rate": 2.3132626525305065e-06, "loss": 0.03, "step": 19226 }, { "epoch": 5.97, "learning_rate": 2.312862572514503e-06, "loss": 0.0012, "step": 19227 }, { "epoch": 5.97, "learning_rate": 2.3124624924985e-06, "loss": 0.0345, "step": 19228 }, { "epoch": 5.97, "learning_rate": 2.3120624124824965e-06, "loss": 0.0017, "step": 19229 }, { "epoch": 5.97, "learning_rate": 2.3116623324664936e-06, "loss": 0.0129, "step": 19230 }, { "epoch": 5.97, "learning_rate": 2.3112622524504903e-06, "loss": 0.016, "step": 19231 }, { "epoch": 5.97, "learning_rate": 2.310862172434487e-06, "loss": 0.0058, "step": 19232 }, { "epoch": 5.97, "learning_rate": 2.310462092418484e-06, "loss": 0.0075, "step": 19233 }, { "epoch": 5.97, "learning_rate": 2.3100620124024804e-06, "loss": 0.0227, "step": 19234 }, { "epoch": 5.97, "learning_rate": 2.3096619323864775e-06, "loss": 0.0128, "step": 19235 }, { "epoch": 5.97, "learning_rate": 2.309261852370474e-06, "loss": 0.0099, "step": 19236 }, { "epoch": 5.97, "learning_rate": 2.308861772354471e-06, "loss": 0.0047, "step": 19237 }, { "epoch": 5.97, "learning_rate": 2.308461692338468e-06, "loss": 0.0105, "step": 19238 }, { "epoch": 5.97, "learning_rate": 2.3080616123224647e-06, "loss": 0.0036, "step": 19239 }, { "epoch": 5.97, "learning_rate": 2.3076615323064618e-06, "loss": 0.0058, "step": 19240 }, { "epoch": 5.97, "learning_rate": 2.307261452290458e-06, "loss": 0.005, "step": 19241 }, { "epoch": 5.97, "learning_rate": 2.306861372274455e-06, "loss": 0.0284, "step": 19242 }, { "epoch": 5.97, "learning_rate": 2.306461292258452e-06, "loss": 0.0268, "step": 19243 }, { "epoch": 5.97, "learning_rate": 2.3060612122424485e-06, "loss": 0.0112, "step": 19244 }, { "epoch": 5.97, "learning_rate": 2.3056611322264456e-06, "loss": 0.0039, "step": 19245 }, { "epoch": 5.97, "learning_rate": 2.3052610522104423e-06, "loss": 0.0012, "step": 19246 }, { "epoch": 5.97, "learning_rate": 2.304860972194439e-06, "loss": 0.0191, "step": 19247 }, { "epoch": 5.97, "learning_rate": 2.3044608921784357e-06, "loss": 0.0091, "step": 19248 }, { "epoch": 5.97, "learning_rate": 2.304060812162433e-06, "loss": 0.0252, "step": 19249 }, { "epoch": 5.97, "learning_rate": 2.3036607321464295e-06, "loss": 0.0036, "step": 19250 }, { "epoch": 5.98, "learning_rate": 2.303260652130426e-06, "loss": 0.002, "step": 19251 }, { "epoch": 5.98, "learning_rate": 2.3028605721144233e-06, "loss": 0.0069, "step": 19252 }, { "epoch": 5.98, "learning_rate": 2.30246049209842e-06, "loss": 0.0066, "step": 19253 }, { "epoch": 5.98, "learning_rate": 2.3020604120824166e-06, "loss": 0.0016, "step": 19254 }, { "epoch": 5.98, "learning_rate": 2.3016603320664133e-06, "loss": 0.0303, "step": 19255 }, { "epoch": 5.98, "learning_rate": 2.30126025205041e-06, "loss": 0.0038, "step": 19256 }, { "epoch": 5.98, "learning_rate": 2.300860172034407e-06, "loss": 0.0204, "step": 19257 }, { "epoch": 5.98, "learning_rate": 2.300460092018404e-06, "loss": 0.0018, "step": 19258 }, { "epoch": 5.98, "learning_rate": 2.3000600120024005e-06, "loss": 0.009, "step": 19259 }, { "epoch": 5.98, "learning_rate": 2.2996599319863976e-06, "loss": 0.0048, "step": 19260 }, { "epoch": 5.98, "learning_rate": 2.2992598519703943e-06, "loss": 0.0121, "step": 19261 }, { "epoch": 5.98, "learning_rate": 2.2988597719543914e-06, "loss": 0.013, "step": 19262 }, { "epoch": 5.98, "learning_rate": 2.2984596919383877e-06, "loss": 0.0021, "step": 19263 }, { "epoch": 5.98, "learning_rate": 2.2980596119223848e-06, "loss": 0.0042, "step": 19264 }, { "epoch": 5.98, "learning_rate": 2.2976595319063815e-06, "loss": 0.0044, "step": 19265 }, { "epoch": 5.98, "learning_rate": 2.297259451890378e-06, "loss": 0.0042, "step": 19266 }, { "epoch": 5.98, "learning_rate": 2.2968593718743753e-06, "loss": 0.0087, "step": 19267 }, { "epoch": 5.98, "learning_rate": 2.296459291858372e-06, "loss": 0.01, "step": 19268 }, { "epoch": 5.98, "learning_rate": 2.2960592118423686e-06, "loss": 0.0016, "step": 19269 }, { "epoch": 5.98, "learning_rate": 2.2956591318263653e-06, "loss": 0.0329, "step": 19270 }, { "epoch": 5.98, "learning_rate": 2.2952590518103624e-06, "loss": 0.0045, "step": 19271 }, { "epoch": 5.98, "learning_rate": 2.294858971794359e-06, "loss": 0.0153, "step": 19272 }, { "epoch": 5.98, "learning_rate": 2.294458891778356e-06, "loss": 0.005, "step": 19273 }, { "epoch": 5.98, "learning_rate": 2.294058811762353e-06, "loss": 0.0231, "step": 19274 }, { "epoch": 5.98, "learning_rate": 2.293658731746349e-06, "loss": 0.0121, "step": 19275 }, { "epoch": 5.98, "learning_rate": 2.2932586517303463e-06, "loss": 0.0031, "step": 19276 }, { "epoch": 5.98, "learning_rate": 2.292858571714343e-06, "loss": 0.0116, "step": 19277 }, { "epoch": 5.98, "learning_rate": 2.2924584916983397e-06, "loss": 0.0018, "step": 19278 }, { "epoch": 5.98, "learning_rate": 2.2920584116823368e-06, "loss": 0.0097, "step": 19279 }, { "epoch": 5.98, "learning_rate": 2.2916583316663335e-06, "loss": 0.0111, "step": 19280 }, { "epoch": 5.98, "learning_rate": 2.2912582516503306e-06, "loss": 0.0031, "step": 19281 }, { "epoch": 5.98, "learning_rate": 2.290858171634327e-06, "loss": 0.0029, "step": 19282 }, { "epoch": 5.99, "learning_rate": 2.290458091618324e-06, "loss": 0.0072, "step": 19283 }, { "epoch": 5.99, "learning_rate": 2.2900580116023206e-06, "loss": 0.0075, "step": 19284 }, { "epoch": 5.99, "learning_rate": 2.2896579315863173e-06, "loss": 0.0049, "step": 19285 }, { "epoch": 5.99, "learning_rate": 2.2892578515703144e-06, "loss": 0.0298, "step": 19286 }, { "epoch": 5.99, "learning_rate": 2.288857771554311e-06, "loss": 0.0045, "step": 19287 }, { "epoch": 5.99, "learning_rate": 2.288457691538308e-06, "loss": 0.0066, "step": 19288 }, { "epoch": 5.99, "learning_rate": 2.2880576115223045e-06, "loss": 0.0099, "step": 19289 }, { "epoch": 5.99, "learning_rate": 2.2876575315063016e-06, "loss": 0.0017, "step": 19290 }, { "epoch": 5.99, "learning_rate": 2.2872574514902983e-06, "loss": 0.0031, "step": 19291 }, { "epoch": 5.99, "learning_rate": 2.286857371474295e-06, "loss": 0.0043, "step": 19292 }, { "epoch": 5.99, "learning_rate": 2.286457291458292e-06, "loss": 0.0035, "step": 19293 }, { "epoch": 5.99, "learning_rate": 2.2860572114422883e-06, "loss": 0.0021, "step": 19294 }, { "epoch": 5.99, "learning_rate": 2.2856571314262854e-06, "loss": 0.0028, "step": 19295 }, { "epoch": 5.99, "learning_rate": 2.285257051410282e-06, "loss": 0.0023, "step": 19296 }, { "epoch": 5.99, "learning_rate": 2.284856971394279e-06, "loss": 0.0119, "step": 19297 }, { "epoch": 5.99, "learning_rate": 2.284456891378276e-06, "loss": 0.0024, "step": 19298 }, { "epoch": 5.99, "learning_rate": 2.2840568113622726e-06, "loss": 0.0055, "step": 19299 }, { "epoch": 5.99, "learning_rate": 2.2836567313462693e-06, "loss": 0.0022, "step": 19300 }, { "epoch": 5.99, "learning_rate": 2.2832566513302664e-06, "loss": 0.0032, "step": 19301 }, { "epoch": 5.99, "learning_rate": 2.282856571314263e-06, "loss": 0.0025, "step": 19302 }, { "epoch": 5.99, "learning_rate": 2.2824564912982598e-06, "loss": 0.0234, "step": 19303 }, { "epoch": 5.99, "learning_rate": 2.2820564112822565e-06, "loss": 0.008, "step": 19304 }, { "epoch": 5.99, "learning_rate": 2.2816563312662536e-06, "loss": 0.0109, "step": 19305 }, { "epoch": 5.99, "learning_rate": 2.2812562512502503e-06, "loss": 0.0038, "step": 19306 }, { "epoch": 5.99, "learning_rate": 2.280856171234247e-06, "loss": 0.0066, "step": 19307 }, { "epoch": 5.99, "learning_rate": 2.280456091218244e-06, "loss": 0.004, "step": 19308 }, { "epoch": 5.99, "learning_rate": 2.2800560112022407e-06, "loss": 0.018, "step": 19309 }, { "epoch": 5.99, "learning_rate": 2.2796559311862374e-06, "loss": 0.0066, "step": 19310 }, { "epoch": 5.99, "learning_rate": 2.279255851170234e-06, "loss": 0.0017, "step": 19311 }, { "epoch": 5.99, "learning_rate": 2.2788557711542312e-06, "loss": 0.0022, "step": 19312 }, { "epoch": 5.99, "learning_rate": 2.278455691138228e-06, "loss": 0.0032, "step": 19313 }, { "epoch": 5.99, "learning_rate": 2.2780556111222246e-06, "loss": 0.0128, "step": 19314 }, { "epoch": 5.99, "learning_rate": 2.2776555311062217e-06, "loss": 0.0162, "step": 19315 }, { "epoch": 6.0, "learning_rate": 2.277255451090218e-06, "loss": 0.0079, "step": 19316 }, { "epoch": 6.0, "learning_rate": 2.276855371074215e-06, "loss": 0.0203, "step": 19317 }, { "epoch": 6.0, "learning_rate": 2.2764552910582118e-06, "loss": 0.0031, "step": 19318 }, { "epoch": 6.0, "learning_rate": 2.2760552110422085e-06, "loss": 0.0067, "step": 19319 }, { "epoch": 6.0, "learning_rate": 2.2756551310262056e-06, "loss": 0.0054, "step": 19320 }, { "epoch": 6.0, "learning_rate": 2.2752550510102022e-06, "loss": 0.0033, "step": 19321 }, { "epoch": 6.0, "learning_rate": 2.2748549709941994e-06, "loss": 0.0091, "step": 19322 }, { "epoch": 6.0, "learning_rate": 2.2744548909781956e-06, "loss": 0.0082, "step": 19323 }, { "epoch": 6.0, "learning_rate": 2.2740548109621927e-06, "loss": 0.0015, "step": 19324 }, { "epoch": 6.0, "learning_rate": 2.2736547309461894e-06, "loss": 0.005, "step": 19325 }, { "epoch": 6.0, "learning_rate": 2.273254650930186e-06, "loss": 0.0042, "step": 19326 }, { "epoch": 6.0, "learning_rate": 2.2728545709141832e-06, "loss": 0.0015, "step": 19327 }, { "epoch": 6.0, "learning_rate": 2.27245449089818e-06, "loss": 0.0208, "step": 19328 }, { "epoch": 6.0, "learning_rate": 2.2720544108821766e-06, "loss": 0.0187, "step": 19329 }, { "epoch": 6.0, "learning_rate": 2.2716543308661733e-06, "loss": 0.0029, "step": 19330 }, { "epoch": 6.0, "learning_rate": 2.2712542508501704e-06, "loss": 0.0151, "step": 19331 }, { "epoch": 6.0, "learning_rate": 2.270854170834167e-06, "loss": 0.0018, "step": 19332 }, { "epoch": 6.0, "learning_rate": 2.2704540908181638e-06, "loss": 0.0112, "step": 19333 }, { "epoch": 6.0, "learning_rate": 2.270054010802161e-06, "loss": 0.0016, "step": 19334 }, { "epoch": 6.0, "learning_rate": 2.269653930786157e-06, "loss": 0.0014, "step": 19335 }, { "epoch": 6.0, "learning_rate": 2.2692538507701542e-06, "loss": 0.0015, "step": 19336 }, { "epoch": 6.0, "learning_rate": 2.268853770754151e-06, "loss": 0.0018, "step": 19337 }, { "epoch": 6.0, "learning_rate": 2.2684536907381476e-06, "loss": 0.0023, "step": 19338 }, { "epoch": 6.0, "learning_rate": 2.2680536107221447e-06, "loss": 0.0216, "step": 19339 }, { "epoch": 6.0, "learning_rate": 2.2676535307061414e-06, "loss": 0.0049, "step": 19340 }, { "epoch": 6.0, "learning_rate": 2.267253450690138e-06, "loss": 0.003, "step": 19341 }, { "epoch": 6.0, "learning_rate": 2.2668533706741348e-06, "loss": 0.0011, "step": 19342 }, { "epoch": 6.0, "learning_rate": 2.266453290658132e-06, "loss": 0.0016, "step": 19343 }, { "epoch": 6.0, "learning_rate": 2.2660532106421286e-06, "loss": 0.0013, "step": 19344 }, { "epoch": 6.0, "learning_rate": 2.2656531306261253e-06, "loss": 0.0012, "step": 19345 }, { "epoch": 6.0, "learning_rate": 2.2652530506101224e-06, "loss": 0.0103, "step": 19346 }, { "epoch": 6.0, "learning_rate": 2.264852970594119e-06, "loss": 0.0019, "step": 19347 }, { "epoch": 6.01, "learning_rate": 2.2644528905781157e-06, "loss": 0.0013, "step": 19348 }, { "epoch": 6.01, "learning_rate": 2.2640528105621124e-06, "loss": 0.0019, "step": 19349 }, { "epoch": 6.01, "learning_rate": 2.2636527305461095e-06, "loss": 0.0026, "step": 19350 }, { "epoch": 6.01, "learning_rate": 2.2632526505301062e-06, "loss": 0.0053, "step": 19351 }, { "epoch": 6.01, "learning_rate": 2.262852570514103e-06, "loss": 0.0025, "step": 19352 }, { "epoch": 6.01, "learning_rate": 2.2624524904981e-06, "loss": 0.0015, "step": 19353 }, { "epoch": 6.01, "learning_rate": 2.2620524104820967e-06, "loss": 0.0018, "step": 19354 }, { "epoch": 6.01, "learning_rate": 2.2616523304660934e-06, "loss": 0.0017, "step": 19355 }, { "epoch": 6.01, "learning_rate": 2.26125225045009e-06, "loss": 0.0029, "step": 19356 }, { "epoch": 6.01, "learning_rate": 2.2608521704340868e-06, "loss": 0.0018, "step": 19357 }, { "epoch": 6.01, "learning_rate": 2.260452090418084e-06, "loss": 0.003, "step": 19358 }, { "epoch": 6.01, "learning_rate": 2.2600520104020806e-06, "loss": 0.003, "step": 19359 }, { "epoch": 6.01, "learning_rate": 2.2596519303860772e-06, "loss": 0.0008, "step": 19360 }, { "epoch": 6.01, "learning_rate": 2.2592518503700744e-06, "loss": 0.0016, "step": 19361 }, { "epoch": 6.01, "learning_rate": 2.258851770354071e-06, "loss": 0.0017, "step": 19362 }, { "epoch": 6.01, "learning_rate": 2.2584516903380677e-06, "loss": 0.0007, "step": 19363 }, { "epoch": 6.01, "learning_rate": 2.2580516103220644e-06, "loss": 0.0013, "step": 19364 }, { "epoch": 6.01, "learning_rate": 2.2576515303060615e-06, "loss": 0.0019, "step": 19365 }, { "epoch": 6.01, "learning_rate": 2.2572514502900582e-06, "loss": 0.0103, "step": 19366 }, { "epoch": 6.01, "learning_rate": 2.256851370274055e-06, "loss": 0.0022, "step": 19367 }, { "epoch": 6.01, "learning_rate": 2.256451290258052e-06, "loss": 0.0038, "step": 19368 }, { "epoch": 6.01, "learning_rate": 2.2560512102420487e-06, "loss": 0.0017, "step": 19369 }, { "epoch": 6.01, "learning_rate": 2.2556511302260454e-06, "loss": 0.0025, "step": 19370 }, { "epoch": 6.01, "learning_rate": 2.255251050210042e-06, "loss": 0.0014, "step": 19371 }, { "epoch": 6.01, "learning_rate": 2.254850970194039e-06, "loss": 0.0045, "step": 19372 }, { "epoch": 6.01, "learning_rate": 2.254450890178036e-06, "loss": 0.0081, "step": 19373 }, { "epoch": 6.01, "learning_rate": 2.2540508101620325e-06, "loss": 0.0008, "step": 19374 }, { "epoch": 6.01, "learning_rate": 2.2536507301460297e-06, "loss": 0.0016, "step": 19375 }, { "epoch": 6.01, "learning_rate": 2.253250650130026e-06, "loss": 0.0013, "step": 19376 }, { "epoch": 6.01, "learning_rate": 2.252850570114023e-06, "loss": 0.0038, "step": 19377 }, { "epoch": 6.01, "learning_rate": 2.2524504900980197e-06, "loss": 0.0017, "step": 19378 }, { "epoch": 6.01, "learning_rate": 2.2520504100820164e-06, "loss": 0.0011, "step": 19379 }, { "epoch": 6.02, "learning_rate": 2.2516503300660135e-06, "loss": 0.0176, "step": 19380 }, { "epoch": 6.02, "learning_rate": 2.25125025005001e-06, "loss": 0.0035, "step": 19381 }, { "epoch": 6.02, "learning_rate": 2.250850170034007e-06, "loss": 0.0014, "step": 19382 }, { "epoch": 6.02, "learning_rate": 2.2504500900180036e-06, "loss": 0.0014, "step": 19383 }, { "epoch": 6.02, "learning_rate": 2.2500500100020007e-06, "loss": 0.0048, "step": 19384 }, { "epoch": 6.02, "learning_rate": 2.2496499299859974e-06, "loss": 0.0022, "step": 19385 }, { "epoch": 6.02, "learning_rate": 2.249249849969994e-06, "loss": 0.0013, "step": 19386 }, { "epoch": 6.02, "learning_rate": 2.248849769953991e-06, "loss": 0.0008, "step": 19387 }, { "epoch": 6.02, "learning_rate": 2.248449689937988e-06, "loss": 0.0045, "step": 19388 }, { "epoch": 6.02, "learning_rate": 2.2480496099219845e-06, "loss": 0.0014, "step": 19389 }, { "epoch": 6.02, "learning_rate": 2.2476495299059812e-06, "loss": 0.0014, "step": 19390 }, { "epoch": 6.02, "learning_rate": 2.2472494498899783e-06, "loss": 0.0089, "step": 19391 }, { "epoch": 6.02, "learning_rate": 2.246849369873975e-06, "loss": 0.0007, "step": 19392 }, { "epoch": 6.02, "learning_rate": 2.2464492898579717e-06, "loss": 0.0024, "step": 19393 }, { "epoch": 6.02, "learning_rate": 2.246049209841969e-06, "loss": 0.0464, "step": 19394 }, { "epoch": 6.02, "learning_rate": 2.245649129825965e-06, "loss": 0.0015, "step": 19395 }, { "epoch": 6.02, "learning_rate": 2.245249049809962e-06, "loss": 0.0032, "step": 19396 }, { "epoch": 6.02, "learning_rate": 2.244848969793959e-06, "loss": 0.0008, "step": 19397 }, { "epoch": 6.02, "learning_rate": 2.2444488897779556e-06, "loss": 0.0027, "step": 19398 }, { "epoch": 6.02, "learning_rate": 2.2440488097619527e-06, "loss": 0.001, "step": 19399 }, { "epoch": 6.02, "learning_rate": 2.2436487297459494e-06, "loss": 0.001, "step": 19400 }, { "epoch": 6.02, "learning_rate": 2.243248649729946e-06, "loss": 0.0067, "step": 19401 }, { "epoch": 6.02, "learning_rate": 2.2428485697139427e-06, "loss": 0.0048, "step": 19402 }, { "epoch": 6.02, "learning_rate": 2.24244848969794e-06, "loss": 0.0098, "step": 19403 }, { "epoch": 6.02, "learning_rate": 2.2420484096819365e-06, "loss": 0.0009, "step": 19404 }, { "epoch": 6.02, "learning_rate": 2.2416483296659332e-06, "loss": 0.0047, "step": 19405 }, { "epoch": 6.02, "learning_rate": 2.2412482496499303e-06, "loss": 0.0015, "step": 19406 }, { "epoch": 6.02, "learning_rate": 2.240848169633927e-06, "loss": 0.0014, "step": 19407 }, { "epoch": 6.02, "learning_rate": 2.2404480896179237e-06, "loss": 0.0033, "step": 19408 }, { "epoch": 6.02, "learning_rate": 2.2400480096019204e-06, "loss": 0.0026, "step": 19409 }, { "epoch": 6.02, "learning_rate": 2.2396479295859175e-06, "loss": 0.002, "step": 19410 }, { "epoch": 6.02, "learning_rate": 2.239247849569914e-06, "loss": 0.0016, "step": 19411 }, { "epoch": 6.03, "learning_rate": 2.238847769553911e-06, "loss": 0.0042, "step": 19412 }, { "epoch": 6.03, "learning_rate": 2.238447689537908e-06, "loss": 0.0018, "step": 19413 }, { "epoch": 6.03, "learning_rate": 2.2380476095219047e-06, "loss": 0.0013, "step": 19414 }, { "epoch": 6.03, "learning_rate": 2.2376475295059013e-06, "loss": 0.0023, "step": 19415 }, { "epoch": 6.03, "learning_rate": 2.2372474494898985e-06, "loss": 0.0071, "step": 19416 }, { "epoch": 6.03, "learning_rate": 2.2368473694738947e-06, "loss": 0.0027, "step": 19417 }, { "epoch": 6.03, "learning_rate": 2.236447289457892e-06, "loss": 0.011, "step": 19418 }, { "epoch": 6.03, "learning_rate": 2.2360472094418885e-06, "loss": 0.0064, "step": 19419 }, { "epoch": 6.03, "learning_rate": 2.235647129425885e-06, "loss": 0.0025, "step": 19420 }, { "epoch": 6.03, "learning_rate": 2.2352470494098823e-06, "loss": 0.0021, "step": 19421 }, { "epoch": 6.03, "learning_rate": 2.234846969393879e-06, "loss": 0.0042, "step": 19422 }, { "epoch": 6.03, "learning_rate": 2.2344468893778757e-06, "loss": 0.0119, "step": 19423 }, { "epoch": 6.03, "learning_rate": 2.2340468093618724e-06, "loss": 0.0037, "step": 19424 }, { "epoch": 6.03, "learning_rate": 2.2336467293458695e-06, "loss": 0.0123, "step": 19425 }, { "epoch": 6.03, "learning_rate": 2.233246649329866e-06, "loss": 0.0027, "step": 19426 }, { "epoch": 6.03, "learning_rate": 2.232846569313863e-06, "loss": 0.0252, "step": 19427 }, { "epoch": 6.03, "learning_rate": 2.23244648929786e-06, "loss": 0.0014, "step": 19428 }, { "epoch": 6.03, "learning_rate": 2.2320464092818566e-06, "loss": 0.0016, "step": 19429 }, { "epoch": 6.03, "learning_rate": 2.2316463292658533e-06, "loss": 0.0027, "step": 19430 }, { "epoch": 6.03, "learning_rate": 2.23124624924985e-06, "loss": 0.0014, "step": 19431 }, { "epoch": 6.03, "learning_rate": 2.230846169233847e-06, "loss": 0.002, "step": 19432 }, { "epoch": 6.03, "learning_rate": 2.230446089217844e-06, "loss": 0.0009, "step": 19433 }, { "epoch": 6.03, "learning_rate": 2.2300460092018405e-06, "loss": 0.0013, "step": 19434 }, { "epoch": 6.03, "learning_rate": 2.2296459291858376e-06, "loss": 0.0144, "step": 19435 }, { "epoch": 6.03, "learning_rate": 2.229245849169834e-06, "loss": 0.002, "step": 19436 }, { "epoch": 6.03, "learning_rate": 2.228845769153831e-06, "loss": 0.0023, "step": 19437 }, { "epoch": 6.03, "learning_rate": 2.2284456891378277e-06, "loss": 0.0022, "step": 19438 }, { "epoch": 6.03, "learning_rate": 2.2280456091218244e-06, "loss": 0.0007, "step": 19439 }, { "epoch": 6.03, "learning_rate": 2.2276455291058215e-06, "loss": 0.0017, "step": 19440 }, { "epoch": 6.03, "learning_rate": 2.227245449089818e-06, "loss": 0.0122, "step": 19441 }, { "epoch": 6.03, "learning_rate": 2.226845369073815e-06, "loss": 0.0024, "step": 19442 }, { "epoch": 6.03, "learning_rate": 2.2264452890578115e-06, "loss": 0.0015, "step": 19443 }, { "epoch": 6.03, "learning_rate": 2.2260452090418086e-06, "loss": 0.0013, "step": 19444 }, { "epoch": 6.04, "learning_rate": 2.2256451290258053e-06, "loss": 0.0022, "step": 19445 }, { "epoch": 6.04, "learning_rate": 2.225245049009802e-06, "loss": 0.0119, "step": 19446 }, { "epoch": 6.04, "learning_rate": 2.224844968993799e-06, "loss": 0.0034, "step": 19447 }, { "epoch": 6.04, "learning_rate": 2.224444888977796e-06, "loss": 0.0012, "step": 19448 }, { "epoch": 6.04, "learning_rate": 2.2240448089617925e-06, "loss": 0.0043, "step": 19449 }, { "epoch": 6.04, "learning_rate": 2.223644728945789e-06, "loss": 0.002, "step": 19450 }, { "epoch": 6.04, "learning_rate": 2.2232446489297863e-06, "loss": 0.0011, "step": 19451 }, { "epoch": 6.04, "learning_rate": 2.222844568913783e-06, "loss": 0.0053, "step": 19452 }, { "epoch": 6.04, "learning_rate": 2.2224444888977797e-06, "loss": 0.0011, "step": 19453 }, { "epoch": 6.04, "learning_rate": 2.2220444088817768e-06, "loss": 0.0326, "step": 19454 }, { "epoch": 6.04, "learning_rate": 2.2216443288657735e-06, "loss": 0.0389, "step": 19455 }, { "epoch": 6.04, "learning_rate": 2.22124424884977e-06, "loss": 0.0008, "step": 19456 }, { "epoch": 6.04, "learning_rate": 2.220844168833767e-06, "loss": 0.0024, "step": 19457 }, { "epoch": 6.04, "learning_rate": 2.2204440888177635e-06, "loss": 0.0043, "step": 19458 }, { "epoch": 6.04, "learning_rate": 2.2200440088017606e-06, "loss": 0.0007, "step": 19459 }, { "epoch": 6.04, "learning_rate": 2.2196439287857573e-06, "loss": 0.0007, "step": 19460 }, { "epoch": 6.04, "learning_rate": 2.219243848769754e-06, "loss": 0.0028, "step": 19461 }, { "epoch": 6.04, "learning_rate": 2.218843768753751e-06, "loss": 0.0079, "step": 19462 }, { "epoch": 6.04, "learning_rate": 2.218443688737748e-06, "loss": 0.0017, "step": 19463 }, { "epoch": 6.04, "learning_rate": 2.2180436087217445e-06, "loss": 0.0009, "step": 19464 }, { "epoch": 6.04, "learning_rate": 2.217643528705741e-06, "loss": 0.0017, "step": 19465 }, { "epoch": 6.04, "learning_rate": 2.2172434486897383e-06, "loss": 0.0025, "step": 19466 }, { "epoch": 6.04, "learning_rate": 2.216843368673735e-06, "loss": 0.0018, "step": 19467 }, { "epoch": 6.04, "learning_rate": 2.2164432886577316e-06, "loss": 0.0052, "step": 19468 }, { "epoch": 6.04, "learning_rate": 2.2160432086417288e-06, "loss": 0.0029, "step": 19469 }, { "epoch": 6.04, "learning_rate": 2.2156431286257254e-06, "loss": 0.0077, "step": 19470 }, { "epoch": 6.04, "learning_rate": 2.215243048609722e-06, "loss": 0.0019, "step": 19471 }, { "epoch": 6.04, "learning_rate": 2.214842968593719e-06, "loss": 0.0041, "step": 19472 }, { "epoch": 6.04, "learning_rate": 2.214442888577716e-06, "loss": 0.0011, "step": 19473 }, { "epoch": 6.04, "learning_rate": 2.2140428085617126e-06, "loss": 0.0013, "step": 19474 }, { "epoch": 6.04, "learning_rate": 2.2136427285457093e-06, "loss": 0.0007, "step": 19475 }, { "epoch": 6.04, "learning_rate": 2.2132426485297064e-06, "loss": 0.0035, "step": 19476 }, { "epoch": 6.05, "learning_rate": 2.2128425685137027e-06, "loss": 0.0013, "step": 19477 }, { "epoch": 6.05, "learning_rate": 2.2124424884976998e-06, "loss": 0.0114, "step": 19478 }, { "epoch": 6.05, "learning_rate": 2.2120424084816965e-06, "loss": 0.0014, "step": 19479 }, { "epoch": 6.05, "learning_rate": 2.211642328465693e-06, "loss": 0.0023, "step": 19480 }, { "epoch": 6.05, "learning_rate": 2.2112422484496903e-06, "loss": 0.0015, "step": 19481 }, { "epoch": 6.05, "learning_rate": 2.210842168433687e-06, "loss": 0.0018, "step": 19482 }, { "epoch": 6.05, "learning_rate": 2.2104420884176836e-06, "loss": 0.0032, "step": 19483 }, { "epoch": 6.05, "learning_rate": 2.2100420084016803e-06, "loss": 0.0622, "step": 19484 }, { "epoch": 6.05, "learning_rate": 2.2096419283856774e-06, "loss": 0.0014, "step": 19485 }, { "epoch": 6.05, "learning_rate": 2.209241848369674e-06, "loss": 0.018, "step": 19486 }, { "epoch": 6.05, "learning_rate": 2.208841768353671e-06, "loss": 0.0027, "step": 19487 }, { "epoch": 6.05, "learning_rate": 2.208441688337668e-06, "loss": 0.0011, "step": 19488 }, { "epoch": 6.05, "learning_rate": 2.2080416083216646e-06, "loss": 0.0283, "step": 19489 }, { "epoch": 6.05, "learning_rate": 2.2076415283056613e-06, "loss": 0.0036, "step": 19490 }, { "epoch": 6.05, "learning_rate": 2.207241448289658e-06, "loss": 0.008, "step": 19491 }, { "epoch": 6.05, "learning_rate": 2.206841368273655e-06, "loss": 0.0012, "step": 19492 }, { "epoch": 6.05, "learning_rate": 2.2064412882576518e-06, "loss": 0.0016, "step": 19493 }, { "epoch": 6.05, "learning_rate": 2.2060412082416485e-06, "loss": 0.0396, "step": 19494 }, { "epoch": 6.05, "learning_rate": 2.2056411282256456e-06, "loss": 0.0024, "step": 19495 }, { "epoch": 6.05, "learning_rate": 2.205241048209642e-06, "loss": 0.0011, "step": 19496 }, { "epoch": 6.05, "learning_rate": 2.204840968193639e-06, "loss": 0.0009, "step": 19497 }, { "epoch": 6.05, "learning_rate": 2.2044408881776356e-06, "loss": 0.0144, "step": 19498 }, { "epoch": 6.05, "learning_rate": 2.2040408081616323e-06, "loss": 0.0051, "step": 19499 }, { "epoch": 6.05, "learning_rate": 2.2036407281456294e-06, "loss": 0.0015, "step": 19500 }, { "epoch": 6.05, "learning_rate": 2.203240648129626e-06, "loss": 0.0011, "step": 19501 }, { "epoch": 6.05, "learning_rate": 2.202840568113623e-06, "loss": 0.0041, "step": 19502 }, { "epoch": 6.05, "learning_rate": 2.2024404880976195e-06, "loss": 0.0049, "step": 19503 }, { "epoch": 6.05, "learning_rate": 2.2020404080816166e-06, "loss": 0.0011, "step": 19504 }, { "epoch": 6.05, "learning_rate": 2.2016403280656133e-06, "loss": 0.0054, "step": 19505 }, { "epoch": 6.05, "learning_rate": 2.20124024804961e-06, "loss": 0.0015, "step": 19506 }, { "epoch": 6.05, "learning_rate": 2.200840168033607e-06, "loss": 0.0017, "step": 19507 }, { "epoch": 6.05, "learning_rate": 2.2004400880176038e-06, "loss": 0.0024, "step": 19508 }, { "epoch": 6.06, "learning_rate": 2.2000400080016004e-06, "loss": 0.0013, "step": 19509 }, { "epoch": 6.06, "learning_rate": 2.199639927985597e-06, "loss": 0.0092, "step": 19510 }, { "epoch": 6.06, "learning_rate": 2.1992398479695942e-06, "loss": 0.002, "step": 19511 }, { "epoch": 6.06, "learning_rate": 2.198839767953591e-06, "loss": 0.0023, "step": 19512 }, { "epoch": 6.06, "learning_rate": 2.1984396879375876e-06, "loss": 0.0041, "step": 19513 }, { "epoch": 6.06, "learning_rate": 2.1980396079215847e-06, "loss": 0.0017, "step": 19514 }, { "epoch": 6.06, "learning_rate": 2.1976395279055814e-06, "loss": 0.0014, "step": 19515 }, { "epoch": 6.06, "learning_rate": 2.197239447889578e-06, "loss": 0.0015, "step": 19516 }, { "epoch": 6.06, "learning_rate": 2.1968393678735748e-06, "loss": 0.0017, "step": 19517 }, { "epoch": 6.06, "learning_rate": 2.1968393678735748e-06, "loss": 0.0154, "step": 19518 }, { "epoch": 6.06, "learning_rate": 2.1964392878575715e-06, "loss": 0.0011, "step": 19519 }, { "epoch": 6.06, "learning_rate": 2.1960392078415686e-06, "loss": 0.0473, "step": 19520 }, { "epoch": 6.06, "learning_rate": 2.1956391278255653e-06, "loss": 0.0017, "step": 19521 }, { "epoch": 6.06, "learning_rate": 2.195239047809562e-06, "loss": 0.0028, "step": 19522 }, { "epoch": 6.06, "learning_rate": 2.194838967793559e-06, "loss": 0.0106, "step": 19523 }, { "epoch": 6.06, "learning_rate": 2.1944388877775557e-06, "loss": 0.007, "step": 19524 }, { "epoch": 6.06, "learning_rate": 2.1940388077615524e-06, "loss": 0.002, "step": 19525 }, { "epoch": 6.06, "learning_rate": 2.193638727745549e-06, "loss": 0.0034, "step": 19526 }, { "epoch": 6.06, "learning_rate": 2.1932386477295462e-06, "loss": 0.0009, "step": 19527 }, { "epoch": 6.06, "learning_rate": 2.192838567713543e-06, "loss": 0.0008, "step": 19528 }, { "epoch": 6.06, "learning_rate": 2.1924384876975396e-06, "loss": 0.0083, "step": 19529 }, { "epoch": 6.06, "learning_rate": 2.1920384076815367e-06, "loss": 0.0017, "step": 19530 }, { "epoch": 6.06, "learning_rate": 2.1916383276655334e-06, "loss": 0.0019, "step": 19531 }, { "epoch": 6.06, "learning_rate": 2.19123824764953e-06, "loss": 0.0029, "step": 19532 }, { "epoch": 6.06, "learning_rate": 2.1908381676335268e-06, "loss": 0.0012, "step": 19533 }, { "epoch": 6.06, "learning_rate": 2.190438087617524e-06, "loss": 0.0047, "step": 19534 }, { "epoch": 6.06, "learning_rate": 2.1900380076015206e-06, "loss": 0.0011, "step": 19535 }, { "epoch": 6.06, "learning_rate": 2.1896379275855173e-06, "loss": 0.0024, "step": 19536 }, { "epoch": 6.06, "learning_rate": 2.1892378475695144e-06, "loss": 0.0011, "step": 19537 }, { "epoch": 6.06, "learning_rate": 2.1888377675535106e-06, "loss": 0.0021, "step": 19538 }, { "epoch": 6.06, "learning_rate": 2.1884376875375077e-06, "loss": 0.0016, "step": 19539 }, { "epoch": 6.06, "learning_rate": 2.1880376075215044e-06, "loss": 0.0024, "step": 19540 }, { "epoch": 6.07, "learning_rate": 2.187637527505501e-06, "loss": 0.0036, "step": 19541 }, { "epoch": 6.07, "learning_rate": 2.1872374474894982e-06, "loss": 0.0012, "step": 19542 }, { "epoch": 6.07, "learning_rate": 2.186837367473495e-06, "loss": 0.0203, "step": 19543 }, { "epoch": 6.07, "learning_rate": 2.1864372874574916e-06, "loss": 0.001, "step": 19544 }, { "epoch": 6.07, "learning_rate": 2.1860372074414883e-06, "loss": 0.0371, "step": 19545 }, { "epoch": 6.07, "learning_rate": 2.1856371274254854e-06, "loss": 0.0069, "step": 19546 }, { "epoch": 6.07, "learning_rate": 2.185237047409482e-06, "loss": 0.0017, "step": 19547 }, { "epoch": 6.07, "learning_rate": 2.1848369673934788e-06, "loss": 0.0015, "step": 19548 }, { "epoch": 6.07, "learning_rate": 2.184436887377476e-06, "loss": 0.0011, "step": 19549 }, { "epoch": 6.07, "learning_rate": 2.1840368073614726e-06, "loss": 0.0043, "step": 19550 }, { "epoch": 6.07, "learning_rate": 2.1836367273454692e-06, "loss": 0.0024, "step": 19551 }, { "epoch": 6.07, "learning_rate": 2.183236647329466e-06, "loss": 0.0019, "step": 19552 }, { "epoch": 6.07, "learning_rate": 2.182836567313463e-06, "loss": 0.0016, "step": 19553 }, { "epoch": 6.07, "learning_rate": 2.1824364872974597e-06, "loss": 0.0045, "step": 19554 }, { "epoch": 6.07, "learning_rate": 2.1820364072814564e-06, "loss": 0.0011, "step": 19555 }, { "epoch": 6.07, "learning_rate": 2.1816363272654535e-06, "loss": 0.0022, "step": 19556 }, { "epoch": 6.07, "learning_rate": 2.1812362472494498e-06, "loss": 0.0029, "step": 19557 }, { "epoch": 6.07, "learning_rate": 2.180836167233447e-06, "loss": 0.0012, "step": 19558 }, { "epoch": 6.07, "learning_rate": 2.1804360872174436e-06, "loss": 0.0109, "step": 19559 }, { "epoch": 6.07, "learning_rate": 2.1800360072014403e-06, "loss": 0.0017, "step": 19560 }, { "epoch": 6.07, "learning_rate": 2.1796359271854374e-06, "loss": 0.001, "step": 19561 }, { "epoch": 6.07, "learning_rate": 2.179235847169434e-06, "loss": 0.0013, "step": 19562 }, { "epoch": 6.07, "learning_rate": 2.1788357671534307e-06, "loss": 0.0024, "step": 19563 }, { "epoch": 6.07, "learning_rate": 2.178435687137428e-06, "loss": 0.0036, "step": 19564 }, { "epoch": 6.07, "learning_rate": 2.1780356071214245e-06, "loss": 0.0312, "step": 19565 }, { "epoch": 6.07, "learning_rate": 2.1776355271054212e-06, "loss": 0.0018, "step": 19566 }, { "epoch": 6.07, "learning_rate": 2.177235447089418e-06, "loss": 0.0012, "step": 19567 }, { "epoch": 6.07, "learning_rate": 2.176835367073415e-06, "loss": 0.0028, "step": 19568 }, { "epoch": 6.07, "learning_rate": 2.1764352870574117e-06, "loss": 0.003, "step": 19569 }, { "epoch": 6.07, "learning_rate": 2.1760352070414084e-06, "loss": 0.0011, "step": 19570 }, { "epoch": 6.07, "learning_rate": 2.1756351270254055e-06, "loss": 0.0018, "step": 19571 }, { "epoch": 6.07, "learning_rate": 2.175235047009402e-06, "loss": 0.0066, "step": 19572 }, { "epoch": 6.08, "learning_rate": 2.174834966993399e-06, "loss": 0.0018, "step": 19573 }, { "epoch": 6.08, "learning_rate": 2.1744348869773956e-06, "loss": 0.0024, "step": 19574 }, { "epoch": 6.08, "learning_rate": 2.1740348069613927e-06, "loss": 0.0015, "step": 19575 }, { "epoch": 6.08, "learning_rate": 2.1736347269453894e-06, "loss": 0.004, "step": 19576 }, { "epoch": 6.08, "learning_rate": 2.173234646929386e-06, "loss": 0.0027, "step": 19577 }, { "epoch": 6.08, "learning_rate": 2.172834566913383e-06, "loss": 0.0019, "step": 19578 }, { "epoch": 6.08, "learning_rate": 2.1724344868973794e-06, "loss": 0.0015, "step": 19579 }, { "epoch": 6.08, "learning_rate": 2.1720344068813765e-06, "loss": 0.007, "step": 19580 }, { "epoch": 6.08, "learning_rate": 2.1716343268653732e-06, "loss": 0.0032, "step": 19581 }, { "epoch": 6.08, "learning_rate": 2.17123424684937e-06, "loss": 0.0013, "step": 19582 }, { "epoch": 6.08, "learning_rate": 2.170834166833367e-06, "loss": 0.0031, "step": 19583 }, { "epoch": 6.08, "learning_rate": 2.1704340868173637e-06, "loss": 0.0069, "step": 19584 }, { "epoch": 6.08, "learning_rate": 2.1700340068013604e-06, "loss": 0.0119, "step": 19585 }, { "epoch": 6.08, "learning_rate": 2.169633926785357e-06, "loss": 0.0013, "step": 19586 }, { "epoch": 6.08, "learning_rate": 2.169233846769354e-06, "loss": 0.0022, "step": 19587 }, { "epoch": 6.08, "learning_rate": 2.168833766753351e-06, "loss": 0.0311, "step": 19588 }, { "epoch": 6.08, "learning_rate": 2.1684336867373476e-06, "loss": 0.0022, "step": 19589 }, { "epoch": 6.08, "learning_rate": 2.1680336067213447e-06, "loss": 0.0083, "step": 19590 }, { "epoch": 6.08, "learning_rate": 2.167633526705341e-06, "loss": 0.0032, "step": 19591 }, { "epoch": 6.08, "learning_rate": 2.167233446689338e-06, "loss": 0.0014, "step": 19592 }, { "epoch": 6.08, "learning_rate": 2.1668333666733347e-06, "loss": 0.0017, "step": 19593 }, { "epoch": 6.08, "learning_rate": 2.166433286657332e-06, "loss": 0.0023, "step": 19594 }, { "epoch": 6.08, "learning_rate": 2.1660332066413285e-06, "loss": 0.0018, "step": 19595 }, { "epoch": 6.08, "learning_rate": 2.165633126625325e-06, "loss": 0.0014, "step": 19596 }, { "epoch": 6.08, "learning_rate": 2.1652330466093223e-06, "loss": 0.0037, "step": 19597 }, { "epoch": 6.08, "learning_rate": 2.1648329665933186e-06, "loss": 0.0101, "step": 19598 }, { "epoch": 6.08, "learning_rate": 2.1644328865773157e-06, "loss": 0.0489, "step": 19599 }, { "epoch": 6.08, "learning_rate": 2.1640328065613124e-06, "loss": 0.0021, "step": 19600 }, { "epoch": 6.08, "learning_rate": 2.163632726545309e-06, "loss": 0.0021, "step": 19601 }, { "epoch": 6.08, "learning_rate": 2.163232646529306e-06, "loss": 0.0023, "step": 19602 }, { "epoch": 6.08, "learning_rate": 2.162832566513303e-06, "loss": 0.0175, "step": 19603 }, { "epoch": 6.08, "learning_rate": 2.1624324864972995e-06, "loss": 0.0025, "step": 19604 }, { "epoch": 6.08, "learning_rate": 2.1620324064812962e-06, "loss": 0.0036, "step": 19605 }, { "epoch": 6.09, "learning_rate": 2.1616323264652933e-06, "loss": 0.0018, "step": 19606 }, { "epoch": 6.09, "learning_rate": 2.16123224644929e-06, "loss": 0.0093, "step": 19607 }, { "epoch": 6.09, "learning_rate": 2.1608321664332867e-06, "loss": 0.0013, "step": 19608 }, { "epoch": 6.09, "learning_rate": 2.160432086417284e-06, "loss": 0.0035, "step": 19609 }, { "epoch": 6.09, "learning_rate": 2.1600320064012805e-06, "loss": 0.0024, "step": 19610 }, { "epoch": 6.09, "learning_rate": 2.159631926385277e-06, "loss": 0.0023, "step": 19611 }, { "epoch": 6.09, "learning_rate": 2.159231846369274e-06, "loss": 0.0013, "step": 19612 }, { "epoch": 6.09, "learning_rate": 2.158831766353271e-06, "loss": 0.0028, "step": 19613 }, { "epoch": 6.09, "learning_rate": 2.1584316863372677e-06, "loss": 0.0016, "step": 19614 }, { "epoch": 6.09, "learning_rate": 2.1580316063212644e-06, "loss": 0.0015, "step": 19615 }, { "epoch": 6.09, "learning_rate": 2.1576315263052615e-06, "loss": 0.0017, "step": 19616 }, { "epoch": 6.09, "learning_rate": 2.157231446289258e-06, "loss": 0.003, "step": 19617 }, { "epoch": 6.09, "learning_rate": 2.156831366273255e-06, "loss": 0.0008, "step": 19618 }, { "epoch": 6.09, "learning_rate": 2.1564312862572515e-06, "loss": 0.001, "step": 19619 }, { "epoch": 6.09, "learning_rate": 2.1560312062412482e-06, "loss": 0.02, "step": 19620 }, { "epoch": 6.09, "learning_rate": 2.1556311262252453e-06, "loss": 0.0058, "step": 19621 }, { "epoch": 6.09, "learning_rate": 2.155231046209242e-06, "loss": 0.0018, "step": 19622 }, { "epoch": 6.09, "learning_rate": 2.1548309661932387e-06, "loss": 0.0071, "step": 19623 }, { "epoch": 6.09, "learning_rate": 2.154430886177236e-06, "loss": 0.0016, "step": 19624 }, { "epoch": 6.09, "learning_rate": 2.1540308061612325e-06, "loss": 0.0014, "step": 19625 }, { "epoch": 6.09, "learning_rate": 2.153630726145229e-06, "loss": 0.0024, "step": 19626 }, { "epoch": 6.09, "learning_rate": 2.153230646129226e-06, "loss": 0.0016, "step": 19627 }, { "epoch": 6.09, "learning_rate": 2.152830566113223e-06, "loss": 0.0022, "step": 19628 }, { "epoch": 6.09, "learning_rate": 2.1524304860972197e-06, "loss": 0.0055, "step": 19629 }, { "epoch": 6.09, "learning_rate": 2.1520304060812163e-06, "loss": 0.0021, "step": 19630 }, { "epoch": 6.09, "learning_rate": 2.1516303260652135e-06, "loss": 0.0035, "step": 19631 }, { "epoch": 6.09, "learning_rate": 2.1512302460492097e-06, "loss": 0.0018, "step": 19632 }, { "epoch": 6.09, "learning_rate": 2.150830166033207e-06, "loss": 0.0027, "step": 19633 }, { "epoch": 6.09, "learning_rate": 2.1504300860172035e-06, "loss": 0.0041, "step": 19634 }, { "epoch": 6.09, "learning_rate": 2.1500300060012006e-06, "loss": 0.0016, "step": 19635 }, { "epoch": 6.09, "learning_rate": 2.1496299259851973e-06, "loss": 0.0074, "step": 19636 }, { "epoch": 6.09, "learning_rate": 2.149229845969194e-06, "loss": 0.0035, "step": 19637 }, { "epoch": 6.1, "learning_rate": 2.148829765953191e-06, "loss": 0.0037, "step": 19638 }, { "epoch": 6.1, "learning_rate": 2.1484296859371874e-06, "loss": 0.0046, "step": 19639 }, { "epoch": 6.1, "learning_rate": 2.1480296059211845e-06, "loss": 0.0009, "step": 19640 }, { "epoch": 6.1, "learning_rate": 2.147629525905181e-06, "loss": 0.0084, "step": 19641 }, { "epoch": 6.1, "learning_rate": 2.147229445889178e-06, "loss": 0.0025, "step": 19642 }, { "epoch": 6.1, "learning_rate": 2.146829365873175e-06, "loss": 0.0023, "step": 19643 }, { "epoch": 6.1, "learning_rate": 2.1464292858571717e-06, "loss": 0.0015, "step": 19644 }, { "epoch": 6.1, "learning_rate": 2.1460292058411683e-06, "loss": 0.0018, "step": 19645 }, { "epoch": 6.1, "learning_rate": 2.145629125825165e-06, "loss": 0.0177, "step": 19646 }, { "epoch": 6.1, "learning_rate": 2.145229045809162e-06, "loss": 0.0015, "step": 19647 }, { "epoch": 6.1, "learning_rate": 2.144828965793159e-06, "loss": 0.0018, "step": 19648 }, { "epoch": 6.1, "learning_rate": 2.1444288857771555e-06, "loss": 0.0017, "step": 19649 }, { "epoch": 6.1, "learning_rate": 2.1440288057611526e-06, "loss": 0.0056, "step": 19650 }, { "epoch": 6.1, "learning_rate": 2.143628725745149e-06, "loss": 0.003, "step": 19651 }, { "epoch": 6.1, "learning_rate": 2.143228645729146e-06, "loss": 0.003, "step": 19652 }, { "epoch": 6.1, "learning_rate": 2.1428285657131427e-06, "loss": 0.002, "step": 19653 }, { "epoch": 6.1, "learning_rate": 2.1424284856971398e-06, "loss": 0.0347, "step": 19654 }, { "epoch": 6.1, "learning_rate": 2.1420284056811365e-06, "loss": 0.0082, "step": 19655 }, { "epoch": 6.1, "learning_rate": 2.141628325665133e-06, "loss": 0.0017, "step": 19656 }, { "epoch": 6.1, "learning_rate": 2.1412282456491303e-06, "loss": 0.001, "step": 19657 }, { "epoch": 6.1, "learning_rate": 2.1408281656331265e-06, "loss": 0.0027, "step": 19658 }, { "epoch": 6.1, "learning_rate": 2.1404280856171236e-06, "loss": 0.0012, "step": 19659 }, { "epoch": 6.1, "learning_rate": 2.1400280056011203e-06, "loss": 0.0227, "step": 19660 }, { "epoch": 6.1, "learning_rate": 2.139627925585117e-06, "loss": 0.0031, "step": 19661 }, { "epoch": 6.1, "learning_rate": 2.139227845569114e-06, "loss": 0.0036, "step": 19662 }, { "epoch": 6.1, "learning_rate": 2.138827765553111e-06, "loss": 0.0009, "step": 19663 }, { "epoch": 6.1, "learning_rate": 2.1384276855371075e-06, "loss": 0.0008, "step": 19664 }, { "epoch": 6.1, "learning_rate": 2.138027605521104e-06, "loss": 0.0028, "step": 19665 }, { "epoch": 6.1, "learning_rate": 2.1376275255051013e-06, "loss": 0.0013, "step": 19666 }, { "epoch": 6.1, "learning_rate": 2.137227445489098e-06, "loss": 0.0013, "step": 19667 }, { "epoch": 6.1, "learning_rate": 2.1368273654730947e-06, "loss": 0.0044, "step": 19668 }, { "epoch": 6.1, "learning_rate": 2.1364272854570918e-06, "loss": 0.0019, "step": 19669 }, { "epoch": 6.11, "learning_rate": 2.1360272054410885e-06, "loss": 0.0018, "step": 19670 }, { "epoch": 6.11, "learning_rate": 2.135627125425085e-06, "loss": 0.0039, "step": 19671 }, { "epoch": 6.11, "learning_rate": 2.1352270454090823e-06, "loss": 0.0012, "step": 19672 }, { "epoch": 6.11, "learning_rate": 2.1348269653930785e-06, "loss": 0.0124, "step": 19673 }, { "epoch": 6.11, "learning_rate": 2.1344268853770756e-06, "loss": 0.0023, "step": 19674 }, { "epoch": 6.11, "learning_rate": 2.1340268053610723e-06, "loss": 0.0088, "step": 19675 }, { "epoch": 6.11, "learning_rate": 2.1336267253450694e-06, "loss": 0.0013, "step": 19676 }, { "epoch": 6.11, "learning_rate": 2.133226645329066e-06, "loss": 0.0024, "step": 19677 }, { "epoch": 6.11, "learning_rate": 2.132826565313063e-06, "loss": 0.0023, "step": 19678 }, { "epoch": 6.11, "learning_rate": 2.13242648529706e-06, "loss": 0.0021, "step": 19679 }, { "epoch": 6.11, "learning_rate": 2.132026405281056e-06, "loss": 0.0018, "step": 19680 }, { "epoch": 6.11, "learning_rate": 2.1316263252650533e-06, "loss": 0.0019, "step": 19681 }, { "epoch": 6.11, "learning_rate": 2.13122624524905e-06, "loss": 0.0027, "step": 19682 }, { "epoch": 6.11, "learning_rate": 2.1308261652330467e-06, "loss": 0.0053, "step": 19683 }, { "epoch": 6.11, "learning_rate": 2.1304260852170438e-06, "loss": 0.001, "step": 19684 }, { "epoch": 6.11, "learning_rate": 2.1300260052010404e-06, "loss": 0.001, "step": 19685 }, { "epoch": 6.11, "learning_rate": 2.129625925185037e-06, "loss": 0.0073, "step": 19686 }, { "epoch": 6.11, "learning_rate": 2.129225845169034e-06, "loss": 0.0015, "step": 19687 }, { "epoch": 6.11, "learning_rate": 2.128825765153031e-06, "loss": 0.0008, "step": 19688 }, { "epoch": 6.11, "learning_rate": 2.1284256851370276e-06, "loss": 0.0021, "step": 19689 }, { "epoch": 6.11, "learning_rate": 2.1280256051210243e-06, "loss": 0.0022, "step": 19690 }, { "epoch": 6.11, "learning_rate": 2.1276255251050214e-06, "loss": 0.0014, "step": 19691 }, { "epoch": 6.11, "learning_rate": 2.1272254450890177e-06, "loss": 0.001, "step": 19692 }, { "epoch": 6.11, "learning_rate": 2.1268253650730148e-06, "loss": 0.0392, "step": 19693 }, { "epoch": 6.11, "learning_rate": 2.1264252850570115e-06, "loss": 0.0695, "step": 19694 }, { "epoch": 6.11, "learning_rate": 2.1260252050410086e-06, "loss": 0.0044, "step": 19695 }, { "epoch": 6.11, "learning_rate": 2.1256251250250053e-06, "loss": 0.0022, "step": 19696 }, { "epoch": 6.11, "learning_rate": 2.125225045009002e-06, "loss": 0.0015, "step": 19697 }, { "epoch": 6.11, "learning_rate": 2.124824964992999e-06, "loss": 0.0037, "step": 19698 }, { "epoch": 6.11, "learning_rate": 2.1244248849769953e-06, "loss": 0.0011, "step": 19699 }, { "epoch": 6.11, "learning_rate": 2.1240248049609924e-06, "loss": 0.0012, "step": 19700 }, { "epoch": 6.11, "learning_rate": 2.123624724944989e-06, "loss": 0.0039, "step": 19701 }, { "epoch": 6.12, "learning_rate": 2.123224644928986e-06, "loss": 0.0055, "step": 19702 }, { "epoch": 6.12, "learning_rate": 2.122824564912983e-06, "loss": 0.0011, "step": 19703 }, { "epoch": 6.12, "learning_rate": 2.1224244848969796e-06, "loss": 0.0108, "step": 19704 }, { "epoch": 6.12, "learning_rate": 2.1220244048809763e-06, "loss": 0.011, "step": 19705 }, { "epoch": 6.12, "learning_rate": 2.121624324864973e-06, "loss": 0.0015, "step": 19706 }, { "epoch": 6.12, "learning_rate": 2.12122424484897e-06, "loss": 0.0009, "step": 19707 }, { "epoch": 6.12, "learning_rate": 2.1208241648329668e-06, "loss": 0.0026, "step": 19708 }, { "epoch": 6.12, "learning_rate": 2.1204240848169635e-06, "loss": 0.0025, "step": 19709 }, { "epoch": 6.12, "learning_rate": 2.1200240048009606e-06, "loss": 0.0019, "step": 19710 }, { "epoch": 6.12, "learning_rate": 2.119623924784957e-06, "loss": 0.0039, "step": 19711 }, { "epoch": 6.12, "learning_rate": 2.119223844768954e-06, "loss": 0.0007, "step": 19712 }, { "epoch": 6.12, "learning_rate": 2.1188237647529506e-06, "loss": 0.0026, "step": 19713 }, { "epoch": 6.12, "learning_rate": 2.1184236847369473e-06, "loss": 0.0019, "step": 19714 }, { "epoch": 6.12, "learning_rate": 2.1180236047209444e-06, "loss": 0.0022, "step": 19715 }, { "epoch": 6.12, "learning_rate": 2.117623524704941e-06, "loss": 0.0018, "step": 19716 }, { "epoch": 6.12, "learning_rate": 2.1172234446889382e-06, "loss": 0.0014, "step": 19717 }, { "epoch": 6.12, "learning_rate": 2.116823364672935e-06, "loss": 0.0031, "step": 19718 }, { "epoch": 6.12, "learning_rate": 2.1164232846569316e-06, "loss": 0.0032, "step": 19719 }, { "epoch": 6.12, "learning_rate": 2.1160232046409283e-06, "loss": 0.0369, "step": 19720 }, { "epoch": 6.12, "learning_rate": 2.115623124624925e-06, "loss": 0.0017, "step": 19721 }, { "epoch": 6.12, "learning_rate": 2.115223044608922e-06, "loss": 0.0025, "step": 19722 }, { "epoch": 6.12, "learning_rate": 2.1148229645929188e-06, "loss": 0.0008, "step": 19723 }, { "epoch": 6.12, "learning_rate": 2.1144228845769154e-06, "loss": 0.0019, "step": 19724 }, { "epoch": 6.12, "learning_rate": 2.1140228045609126e-06, "loss": 0.0017, "step": 19725 }, { "epoch": 6.12, "learning_rate": 2.1136227245449092e-06, "loss": 0.0122, "step": 19726 }, { "epoch": 6.12, "learning_rate": 2.113222644528906e-06, "loss": 0.0016, "step": 19727 }, { "epoch": 6.12, "learning_rate": 2.1128225645129026e-06, "loss": 0.0026, "step": 19728 }, { "epoch": 6.12, "learning_rate": 2.1124224844968997e-06, "loss": 0.0026, "step": 19729 }, { "epoch": 6.12, "learning_rate": 2.1120224044808964e-06, "loss": 0.0073, "step": 19730 }, { "epoch": 6.12, "learning_rate": 2.111622324464893e-06, "loss": 0.0012, "step": 19731 }, { "epoch": 6.12, "learning_rate": 2.11122224444889e-06, "loss": 0.0016, "step": 19732 }, { "epoch": 6.12, "learning_rate": 2.1108221644328865e-06, "loss": 0.0028, "step": 19733 }, { "epoch": 6.13, "learning_rate": 2.1104220844168836e-06, "loss": 0.0039, "step": 19734 }, { "epoch": 6.13, "learning_rate": 2.1100220044008803e-06, "loss": 0.0037, "step": 19735 }, { "epoch": 6.13, "learning_rate": 2.1096219243848774e-06, "loss": 0.0007, "step": 19736 }, { "epoch": 6.13, "learning_rate": 2.109221844368874e-06, "loss": 0.0018, "step": 19737 }, { "epoch": 6.13, "learning_rate": 2.1088217643528707e-06, "loss": 0.0013, "step": 19738 }, { "epoch": 6.13, "learning_rate": 2.108421684336868e-06, "loss": 0.0016, "step": 19739 }, { "epoch": 6.13, "learning_rate": 2.108021604320864e-06, "loss": 0.0009, "step": 19740 }, { "epoch": 6.13, "learning_rate": 2.1076215243048612e-06, "loss": 0.0042, "step": 19741 }, { "epoch": 6.13, "learning_rate": 2.107221444288858e-06, "loss": 0.0019, "step": 19742 }, { "epoch": 6.13, "learning_rate": 2.1068213642728546e-06, "loss": 0.0018, "step": 19743 }, { "epoch": 6.13, "learning_rate": 2.1064212842568517e-06, "loss": 0.0032, "step": 19744 }, { "epoch": 6.13, "learning_rate": 2.1060212042408484e-06, "loss": 0.0015, "step": 19745 }, { "epoch": 6.13, "learning_rate": 2.105621124224845e-06, "loss": 0.0015, "step": 19746 }, { "epoch": 6.13, "learning_rate": 2.1052210442088418e-06, "loss": 0.0027, "step": 19747 }, { "epoch": 6.13, "learning_rate": 2.104820964192839e-06, "loss": 0.0014, "step": 19748 }, { "epoch": 6.13, "learning_rate": 2.1044208841768356e-06, "loss": 0.0011, "step": 19749 }, { "epoch": 6.13, "learning_rate": 2.1040208041608323e-06, "loss": 0.0025, "step": 19750 }, { "epoch": 6.13, "learning_rate": 2.1036207241448294e-06, "loss": 0.0018, "step": 19751 }, { "epoch": 6.13, "learning_rate": 2.1032206441288256e-06, "loss": 0.0008, "step": 19752 }, { "epoch": 6.13, "learning_rate": 2.1028205641128227e-06, "loss": 0.0201, "step": 19753 }, { "epoch": 6.13, "learning_rate": 2.1024204840968194e-06, "loss": 0.0057, "step": 19754 }, { "epoch": 6.13, "learning_rate": 2.102020404080816e-06, "loss": 0.0125, "step": 19755 }, { "epoch": 6.13, "learning_rate": 2.1016203240648132e-06, "loss": 0.0018, "step": 19756 }, { "epoch": 6.13, "learning_rate": 2.10122024404881e-06, "loss": 0.0037, "step": 19757 }, { "epoch": 6.13, "learning_rate": 2.100820164032807e-06, "loss": 0.0014, "step": 19758 }, { "epoch": 6.13, "learning_rate": 2.1004200840168033e-06, "loss": 0.0015, "step": 19759 }, { "epoch": 6.13, "learning_rate": 2.1000200040008004e-06, "loss": 0.0012, "step": 19760 }, { "epoch": 6.13, "learning_rate": 2.099619923984797e-06, "loss": 0.003, "step": 19761 }, { "epoch": 6.13, "learning_rate": 2.0992198439687938e-06, "loss": 0.0022, "step": 19762 }, { "epoch": 6.13, "learning_rate": 2.098819763952791e-06, "loss": 0.0318, "step": 19763 }, { "epoch": 6.13, "learning_rate": 2.0984196839367876e-06, "loss": 0.0019, "step": 19764 }, { "epoch": 6.13, "learning_rate": 2.0980196039207842e-06, "loss": 0.0026, "step": 19765 }, { "epoch": 6.13, "learning_rate": 2.097619523904781e-06, "loss": 0.0165, "step": 19766 }, { "epoch": 6.14, "learning_rate": 2.097219443888778e-06, "loss": 0.001, "step": 19767 }, { "epoch": 6.14, "learning_rate": 2.0968193638727747e-06, "loss": 0.0041, "step": 19768 }, { "epoch": 6.14, "learning_rate": 2.0964192838567714e-06, "loss": 0.0028, "step": 19769 }, { "epoch": 6.14, "learning_rate": 2.0960192038407685e-06, "loss": 0.0015, "step": 19770 }, { "epoch": 6.14, "learning_rate": 2.095619123824765e-06, "loss": 0.0037, "step": 19771 }, { "epoch": 6.14, "learning_rate": 2.095219043808762e-06, "loss": 0.0035, "step": 19772 }, { "epoch": 6.14, "learning_rate": 2.0948189637927586e-06, "loss": 0.004, "step": 19773 }, { "epoch": 6.14, "learning_rate": 2.0944188837767553e-06, "loss": 0.0014, "step": 19774 }, { "epoch": 6.14, "learning_rate": 2.0940188037607524e-06, "loss": 0.0014, "step": 19775 }, { "epoch": 6.14, "learning_rate": 2.093618723744749e-06, "loss": 0.0149, "step": 19776 }, { "epoch": 6.14, "learning_rate": 2.093218643728746e-06, "loss": 0.0078, "step": 19777 }, { "epoch": 6.14, "learning_rate": 2.092818563712743e-06, "loss": 0.0024, "step": 19778 }, { "epoch": 6.14, "learning_rate": 2.0924184836967395e-06, "loss": 0.0055, "step": 19779 }, { "epoch": 6.14, "learning_rate": 2.0920184036807362e-06, "loss": 0.0013, "step": 19780 }, { "epoch": 6.14, "learning_rate": 2.091618323664733e-06, "loss": 0.0026, "step": 19781 }, { "epoch": 6.14, "learning_rate": 2.09121824364873e-06, "loss": 0.0047, "step": 19782 }, { "epoch": 6.14, "learning_rate": 2.0908181636327267e-06, "loss": 0.0011, "step": 19783 }, { "epoch": 6.14, "learning_rate": 2.0904180836167234e-06, "loss": 0.0015, "step": 19784 }, { "epoch": 6.14, "learning_rate": 2.0900180036007205e-06, "loss": 0.0019, "step": 19785 }, { "epoch": 6.14, "learning_rate": 2.089617923584717e-06, "loss": 0.0007, "step": 19786 }, { "epoch": 6.14, "learning_rate": 2.089217843568714e-06, "loss": 0.0184, "step": 19787 }, { "epoch": 6.14, "learning_rate": 2.0888177635527106e-06, "loss": 0.003, "step": 19788 }, { "epoch": 6.14, "learning_rate": 2.0884176835367077e-06, "loss": 0.0018, "step": 19789 }, { "epoch": 6.14, "learning_rate": 2.0880176035207044e-06, "loss": 0.0042, "step": 19790 }, { "epoch": 6.14, "learning_rate": 2.087617523504701e-06, "loss": 0.0017, "step": 19791 }, { "epoch": 6.14, "learning_rate": 2.087217443488698e-06, "loss": 0.0013, "step": 19792 }, { "epoch": 6.14, "learning_rate": 2.0868173634726944e-06, "loss": 0.0008, "step": 19793 }, { "epoch": 6.14, "learning_rate": 2.0864172834566915e-06, "loss": 0.0016, "step": 19794 }, { "epoch": 6.14, "learning_rate": 2.0860172034406882e-06, "loss": 0.0091, "step": 19795 }, { "epoch": 6.14, "learning_rate": 2.085617123424685e-06, "loss": 0.0105, "step": 19796 }, { "epoch": 6.14, "learning_rate": 2.085217043408682e-06, "loss": 0.0041, "step": 19797 }, { "epoch": 6.14, "learning_rate": 2.0848169633926787e-06, "loss": 0.0016, "step": 19798 }, { "epoch": 6.15, "learning_rate": 2.084416883376676e-06, "loss": 0.0024, "step": 19799 }, { "epoch": 6.15, "learning_rate": 2.084016803360672e-06, "loss": 0.0015, "step": 19800 }, { "epoch": 6.15, "learning_rate": 2.083616723344669e-06, "loss": 0.0006, "step": 19801 }, { "epoch": 6.15, "learning_rate": 2.083216643328666e-06, "loss": 0.0021, "step": 19802 }, { "epoch": 6.15, "learning_rate": 2.0828165633126626e-06, "loss": 0.0016, "step": 19803 }, { "epoch": 6.15, "learning_rate": 2.0824164832966597e-06, "loss": 0.0041, "step": 19804 }, { "epoch": 6.15, "learning_rate": 2.0820164032806564e-06, "loss": 0.0017, "step": 19805 }, { "epoch": 6.15, "learning_rate": 2.081616323264653e-06, "loss": 0.0013, "step": 19806 }, { "epoch": 6.15, "learning_rate": 2.0812162432486497e-06, "loss": 0.0541, "step": 19807 }, { "epoch": 6.15, "learning_rate": 2.080816163232647e-06, "loss": 0.0018, "step": 19808 }, { "epoch": 6.15, "learning_rate": 2.0804160832166435e-06, "loss": 0.0016, "step": 19809 }, { "epoch": 6.15, "learning_rate": 2.08001600320064e-06, "loss": 0.0012, "step": 19810 }, { "epoch": 6.15, "learning_rate": 2.0796159231846373e-06, "loss": 0.0025, "step": 19811 }, { "epoch": 6.15, "learning_rate": 2.0792158431686336e-06, "loss": 0.0068, "step": 19812 }, { "epoch": 6.15, "learning_rate": 2.0788157631526307e-06, "loss": 0.005, "step": 19813 }, { "epoch": 6.15, "learning_rate": 2.0784156831366274e-06, "loss": 0.0027, "step": 19814 }, { "epoch": 6.15, "learning_rate": 2.078015603120624e-06, "loss": 0.002, "step": 19815 }, { "epoch": 6.15, "learning_rate": 2.077615523104621e-06, "loss": 0.0232, "step": 19816 }, { "epoch": 6.15, "learning_rate": 2.077215443088618e-06, "loss": 0.0015, "step": 19817 }, { "epoch": 6.15, "learning_rate": 2.0768153630726145e-06, "loss": 0.0011, "step": 19818 }, { "epoch": 6.15, "learning_rate": 2.0764152830566112e-06, "loss": 0.0026, "step": 19819 }, { "epoch": 6.15, "learning_rate": 2.0760152030406083e-06, "loss": 0.0032, "step": 19820 }, { "epoch": 6.15, "learning_rate": 2.075615123024605e-06, "loss": 0.0021, "step": 19821 }, { "epoch": 6.15, "learning_rate": 2.0752150430086017e-06, "loss": 0.0007, "step": 19822 }, { "epoch": 6.15, "learning_rate": 2.074814962992599e-06, "loss": 0.0028, "step": 19823 }, { "epoch": 6.15, "learning_rate": 2.0744148829765955e-06, "loss": 0.0014, "step": 19824 }, { "epoch": 6.15, "learning_rate": 2.074014802960592e-06, "loss": 0.003, "step": 19825 }, { "epoch": 6.15, "learning_rate": 2.0736147229445893e-06, "loss": 0.0034, "step": 19826 }, { "epoch": 6.15, "learning_rate": 2.073214642928586e-06, "loss": 0.006, "step": 19827 }, { "epoch": 6.15, "learning_rate": 2.0728145629125827e-06, "loss": 0.0017, "step": 19828 }, { "epoch": 6.15, "learning_rate": 2.0724144828965794e-06, "loss": 0.0037, "step": 19829 }, { "epoch": 6.15, "learning_rate": 2.0720144028805765e-06, "loss": 0.0014, "step": 19830 }, { "epoch": 6.16, "learning_rate": 2.071614322864573e-06, "loss": 0.0016, "step": 19831 }, { "epoch": 6.16, "learning_rate": 2.07121424284857e-06, "loss": 0.0023, "step": 19832 }, { "epoch": 6.16, "learning_rate": 2.070814162832567e-06, "loss": 0.0013, "step": 19833 }, { "epoch": 6.16, "learning_rate": 2.0704140828165632e-06, "loss": 0.001, "step": 19834 }, { "epoch": 6.16, "learning_rate": 2.0700140028005603e-06, "loss": 0.002, "step": 19835 }, { "epoch": 6.16, "learning_rate": 2.069613922784557e-06, "loss": 0.0015, "step": 19836 }, { "epoch": 6.16, "learning_rate": 2.0692138427685537e-06, "loss": 0.0023, "step": 19837 }, { "epoch": 6.16, "learning_rate": 2.068813762752551e-06, "loss": 0.0158, "step": 19838 }, { "epoch": 6.16, "learning_rate": 2.0684136827365475e-06, "loss": 0.0017, "step": 19839 }, { "epoch": 6.16, "learning_rate": 2.0680136027205446e-06, "loss": 0.0013, "step": 19840 }, { "epoch": 6.16, "learning_rate": 2.067613522704541e-06, "loss": 0.0058, "step": 19841 }, { "epoch": 6.16, "learning_rate": 2.067213442688538e-06, "loss": 0.0085, "step": 19842 }, { "epoch": 6.16, "learning_rate": 2.0668133626725347e-06, "loss": 0.0035, "step": 19843 }, { "epoch": 6.16, "learning_rate": 2.0664132826565314e-06, "loss": 0.0018, "step": 19844 }, { "epoch": 6.16, "learning_rate": 2.0660132026405285e-06, "loss": 0.0034, "step": 19845 }, { "epoch": 6.16, "learning_rate": 2.065613122624525e-06, "loss": 0.0007, "step": 19846 }, { "epoch": 6.16, "learning_rate": 2.065213042608522e-06, "loss": 0.0058, "step": 19847 }, { "epoch": 6.16, "learning_rate": 2.0648129625925185e-06, "loss": 0.0009, "step": 19848 }, { "epoch": 6.16, "learning_rate": 2.0644128825765156e-06, "loss": 0.002, "step": 19849 }, { "epoch": 6.16, "learning_rate": 2.0640128025605123e-06, "loss": 0.0012, "step": 19850 }, { "epoch": 6.16, "learning_rate": 2.063612722544509e-06, "loss": 0.0006, "step": 19851 }, { "epoch": 6.16, "learning_rate": 2.063212642528506e-06, "loss": 0.0013, "step": 19852 }, { "epoch": 6.16, "learning_rate": 2.0628125625125024e-06, "loss": 0.0028, "step": 19853 }, { "epoch": 6.16, "learning_rate": 2.0624124824964995e-06, "loss": 0.001, "step": 19854 }, { "epoch": 6.16, "learning_rate": 2.062012402480496e-06, "loss": 0.005, "step": 19855 }, { "epoch": 6.16, "learning_rate": 2.061612322464493e-06, "loss": 0.0024, "step": 19856 }, { "epoch": 6.16, "learning_rate": 2.06121224244849e-06, "loss": 0.0134, "step": 19857 }, { "epoch": 6.16, "learning_rate": 2.0608121624324867e-06, "loss": 0.0028, "step": 19858 }, { "epoch": 6.16, "learning_rate": 2.0604120824164833e-06, "loss": 0.0143, "step": 19859 }, { "epoch": 6.16, "learning_rate": 2.06001200240048e-06, "loss": 0.0008, "step": 19860 }, { "epoch": 6.16, "learning_rate": 2.059611922384477e-06, "loss": 0.0078, "step": 19861 }, { "epoch": 6.16, "learning_rate": 2.059211842368474e-06, "loss": 0.0135, "step": 19862 }, { "epoch": 6.17, "learning_rate": 2.0588117623524705e-06, "loss": 0.0083, "step": 19863 }, { "epoch": 6.17, "learning_rate": 2.0584116823364676e-06, "loss": 0.0013, "step": 19864 }, { "epoch": 6.17, "learning_rate": 2.0580116023204643e-06, "loss": 0.0023, "step": 19865 }, { "epoch": 6.17, "learning_rate": 2.057611522304461e-06, "loss": 0.0013, "step": 19866 }, { "epoch": 6.17, "learning_rate": 2.0572114422884577e-06, "loss": 0.0076, "step": 19867 }, { "epoch": 6.17, "learning_rate": 2.0568113622724548e-06, "loss": 0.0027, "step": 19868 }, { "epoch": 6.17, "learning_rate": 2.0564112822564515e-06, "loss": 0.0113, "step": 19869 }, { "epoch": 6.17, "learning_rate": 2.056011202240448e-06, "loss": 0.0127, "step": 19870 }, { "epoch": 6.17, "learning_rate": 2.0556111222244453e-06, "loss": 0.0016, "step": 19871 }, { "epoch": 6.17, "learning_rate": 2.055211042208442e-06, "loss": 0.0032, "step": 19872 }, { "epoch": 6.17, "learning_rate": 2.0548109621924386e-06, "loss": 0.0015, "step": 19873 }, { "epoch": 6.17, "learning_rate": 2.0544108821764353e-06, "loss": 0.0015, "step": 19874 }, { "epoch": 6.17, "learning_rate": 2.054010802160432e-06, "loss": 0.0028, "step": 19875 }, { "epoch": 6.17, "learning_rate": 2.053610722144429e-06, "loss": 0.0062, "step": 19876 }, { "epoch": 6.17, "learning_rate": 2.053210642128426e-06, "loss": 0.0006, "step": 19877 }, { "epoch": 6.17, "learning_rate": 2.0528105621124225e-06, "loss": 0.025, "step": 19878 }, { "epoch": 6.17, "learning_rate": 2.0524104820964196e-06, "loss": 0.0009, "step": 19879 }, { "epoch": 6.17, "learning_rate": 2.0520104020804163e-06, "loss": 0.0022, "step": 19880 }, { "epoch": 6.17, "learning_rate": 2.051610322064413e-06, "loss": 0.0017, "step": 19881 }, { "epoch": 6.17, "learning_rate": 2.0512102420484097e-06, "loss": 0.0012, "step": 19882 }, { "epoch": 6.17, "learning_rate": 2.0508101620324068e-06, "loss": 0.0179, "step": 19883 }, { "epoch": 6.17, "learning_rate": 2.0504100820164035e-06, "loss": 0.0037, "step": 19884 }, { "epoch": 6.17, "learning_rate": 2.0500100020004e-06, "loss": 0.0029, "step": 19885 }, { "epoch": 6.17, "learning_rate": 2.0496099219843973e-06, "loss": 0.0011, "step": 19886 }, { "epoch": 6.17, "learning_rate": 2.049209841968394e-06, "loss": 0.0008, "step": 19887 }, { "epoch": 6.17, "learning_rate": 2.0488097619523906e-06, "loss": 0.0024, "step": 19888 }, { "epoch": 6.17, "learning_rate": 2.0484096819363873e-06, "loss": 0.0031, "step": 19889 }, { "epoch": 6.17, "learning_rate": 2.0480096019203844e-06, "loss": 0.0026, "step": 19890 }, { "epoch": 6.17, "learning_rate": 2.047609521904381e-06, "loss": 0.0175, "step": 19891 }, { "epoch": 6.17, "learning_rate": 2.047209441888378e-06, "loss": 0.0013, "step": 19892 }, { "epoch": 6.17, "learning_rate": 2.046809361872375e-06, "loss": 0.0006, "step": 19893 }, { "epoch": 6.17, "learning_rate": 2.046409281856371e-06, "loss": 0.0011, "step": 19894 }, { "epoch": 6.17, "learning_rate": 2.0460092018403683e-06, "loss": 0.0014, "step": 19895 }, { "epoch": 6.18, "learning_rate": 2.045609121824365e-06, "loss": 0.001, "step": 19896 }, { "epoch": 6.18, "learning_rate": 2.0452090418083617e-06, "loss": 0.0038, "step": 19897 }, { "epoch": 6.18, "learning_rate": 2.0448089617923588e-06, "loss": 0.0018, "step": 19898 }, { "epoch": 6.18, "learning_rate": 2.0444088817763555e-06, "loss": 0.005, "step": 19899 }, { "epoch": 6.18, "learning_rate": 2.044008801760352e-06, "loss": 0.0045, "step": 19900 }, { "epoch": 6.18, "learning_rate": 2.043608721744349e-06, "loss": 0.0282, "step": 19901 }, { "epoch": 6.18, "learning_rate": 2.043208641728346e-06, "loss": 0.0033, "step": 19902 }, { "epoch": 6.18, "learning_rate": 2.0428085617123426e-06, "loss": 0.0056, "step": 19903 }, { "epoch": 6.18, "learning_rate": 2.0424084816963393e-06, "loss": 0.001, "step": 19904 }, { "epoch": 6.18, "learning_rate": 2.0420084016803364e-06, "loss": 0.0026, "step": 19905 }, { "epoch": 6.18, "learning_rate": 2.041608321664333e-06, "loss": 0.0029, "step": 19906 }, { "epoch": 6.18, "learning_rate": 2.0412082416483298e-06, "loss": 0.0011, "step": 19907 }, { "epoch": 6.18, "learning_rate": 2.0408081616323265e-06, "loss": 0.0166, "step": 19908 }, { "epoch": 6.18, "learning_rate": 2.0404080816163236e-06, "loss": 0.0016, "step": 19909 }, { "epoch": 6.18, "learning_rate": 2.0400080016003203e-06, "loss": 0.0011, "step": 19910 }, { "epoch": 6.18, "learning_rate": 2.039607921584317e-06, "loss": 0.0017, "step": 19911 }, { "epoch": 6.18, "learning_rate": 2.039207841568314e-06, "loss": 0.0104, "step": 19912 }, { "epoch": 6.18, "learning_rate": 2.0388077615523103e-06, "loss": 0.0305, "step": 19913 }, { "epoch": 6.18, "learning_rate": 2.0384076815363074e-06, "loss": 0.0059, "step": 19914 }, { "epoch": 6.18, "learning_rate": 2.038007601520304e-06, "loss": 0.0283, "step": 19915 }, { "epoch": 6.18, "learning_rate": 2.037607521504301e-06, "loss": 0.002, "step": 19916 }, { "epoch": 6.18, "learning_rate": 2.037207441488298e-06, "loss": 0.0271, "step": 19917 }, { "epoch": 6.18, "learning_rate": 2.0368073614722946e-06, "loss": 0.0048, "step": 19918 }, { "epoch": 6.18, "learning_rate": 2.0364072814562913e-06, "loss": 0.0013, "step": 19919 }, { "epoch": 6.18, "learning_rate": 2.036007201440288e-06, "loss": 0.002, "step": 19920 }, { "epoch": 6.18, "learning_rate": 2.035607121424285e-06, "loss": 0.0158, "step": 19921 }, { "epoch": 6.18, "learning_rate": 2.0352070414082818e-06, "loss": 0.0013, "step": 19922 }, { "epoch": 6.18, "learning_rate": 2.0348069613922785e-06, "loss": 0.0022, "step": 19923 }, { "epoch": 6.18, "learning_rate": 2.0344068813762756e-06, "loss": 0.0018, "step": 19924 }, { "epoch": 6.18, "learning_rate": 2.0340068013602723e-06, "loss": 0.0009, "step": 19925 }, { "epoch": 6.18, "learning_rate": 2.033606721344269e-06, "loss": 0.0039, "step": 19926 }, { "epoch": 6.18, "learning_rate": 2.0332066413282656e-06, "loss": 0.0029, "step": 19927 }, { "epoch": 6.19, "learning_rate": 2.0328065613122627e-06, "loss": 0.0037, "step": 19928 }, { "epoch": 6.19, "learning_rate": 2.0324064812962594e-06, "loss": 0.0029, "step": 19929 }, { "epoch": 6.19, "learning_rate": 2.032006401280256e-06, "loss": 0.0011, "step": 19930 }, { "epoch": 6.19, "learning_rate": 2.0316063212642532e-06, "loss": 0.0013, "step": 19931 }, { "epoch": 6.19, "learning_rate": 2.03120624124825e-06, "loss": 0.0367, "step": 19932 }, { "epoch": 6.19, "learning_rate": 2.0308061612322466e-06, "loss": 0.0007, "step": 19933 }, { "epoch": 6.19, "learning_rate": 2.0304060812162437e-06, "loss": 0.0021, "step": 19934 }, { "epoch": 6.19, "learning_rate": 2.03000600120024e-06, "loss": 0.0023, "step": 19935 }, { "epoch": 6.19, "learning_rate": 2.029605921184237e-06, "loss": 0.0019, "step": 19936 }, { "epoch": 6.19, "learning_rate": 2.0292058411682338e-06, "loss": 0.0013, "step": 19937 }, { "epoch": 6.19, "learning_rate": 2.0288057611522305e-06, "loss": 0.004, "step": 19938 }, { "epoch": 6.19, "learning_rate": 2.0284056811362276e-06, "loss": 0.0014, "step": 19939 }, { "epoch": 6.19, "learning_rate": 2.0280056011202242e-06, "loss": 0.0079, "step": 19940 }, { "epoch": 6.19, "learning_rate": 2.027605521104221e-06, "loss": 0.0026, "step": 19941 }, { "epoch": 6.19, "learning_rate": 2.0272054410882176e-06, "loss": 0.0012, "step": 19942 }, { "epoch": 6.19, "learning_rate": 2.0268053610722147e-06, "loss": 0.0032, "step": 19943 }, { "epoch": 6.19, "learning_rate": 2.0264052810562114e-06, "loss": 0.0022, "step": 19944 }, { "epoch": 6.19, "learning_rate": 2.026005201040208e-06, "loss": 0.0007, "step": 19945 }, { "epoch": 6.19, "learning_rate": 2.025605121024205e-06, "loss": 0.0034, "step": 19946 }, { "epoch": 6.19, "learning_rate": 2.025205041008202e-06, "loss": 0.0022, "step": 19947 }, { "epoch": 6.19, "learning_rate": 2.0248049609921986e-06, "loss": 0.0062, "step": 19948 }, { "epoch": 6.19, "learning_rate": 2.0244048809761953e-06, "loss": 0.0365, "step": 19949 }, { "epoch": 6.19, "learning_rate": 2.0240048009601924e-06, "loss": 0.0013, "step": 19950 }, { "epoch": 6.19, "learning_rate": 2.023604720944189e-06, "loss": 0.0012, "step": 19951 }, { "epoch": 6.19, "learning_rate": 2.0232046409281858e-06, "loss": 0.0017, "step": 19952 }, { "epoch": 6.19, "learning_rate": 2.022804560912183e-06, "loss": 0.0023, "step": 19953 }, { "epoch": 6.19, "learning_rate": 2.022404480896179e-06, "loss": 0.0012, "step": 19954 }, { "epoch": 6.19, "learning_rate": 2.0220044008801762e-06, "loss": 0.0047, "step": 19955 }, { "epoch": 6.19, "learning_rate": 2.021604320864173e-06, "loss": 0.003, "step": 19956 }, { "epoch": 6.19, "learning_rate": 2.0212042408481696e-06, "loss": 0.0015, "step": 19957 }, { "epoch": 6.19, "learning_rate": 2.0208041608321667e-06, "loss": 0.002, "step": 19958 }, { "epoch": 6.19, "learning_rate": 2.0204040808161634e-06, "loss": 0.0016, "step": 19959 }, { "epoch": 6.2, "learning_rate": 2.02000400080016e-06, "loss": 0.0008, "step": 19960 }, { "epoch": 6.2, "learning_rate": 2.0196039207841568e-06, "loss": 0.0015, "step": 19961 }, { "epoch": 6.2, "learning_rate": 2.019203840768154e-06, "loss": 0.0011, "step": 19962 }, { "epoch": 6.2, "learning_rate": 2.0188037607521506e-06, "loss": 0.0037, "step": 19963 }, { "epoch": 6.2, "learning_rate": 2.0184036807361473e-06, "loss": 0.0015, "step": 19964 }, { "epoch": 6.2, "learning_rate": 2.0180036007201444e-06, "loss": 0.0401, "step": 19965 }, { "epoch": 6.2, "learning_rate": 2.017603520704141e-06, "loss": 0.0016, "step": 19966 }, { "epoch": 6.2, "learning_rate": 2.0172034406881377e-06, "loss": 0.0018, "step": 19967 }, { "epoch": 6.2, "learning_rate": 2.0168033606721344e-06, "loss": 0.0011, "step": 19968 }, { "epoch": 6.2, "learning_rate": 2.0164032806561315e-06, "loss": 0.0045, "step": 19969 }, { "epoch": 6.2, "learning_rate": 2.0160032006401282e-06, "loss": 0.0867, "step": 19970 }, { "epoch": 6.2, "learning_rate": 2.015603120624125e-06, "loss": 0.0016, "step": 19971 }, { "epoch": 6.2, "learning_rate": 2.015203040608122e-06, "loss": 0.003, "step": 19972 }, { "epoch": 6.2, "learning_rate": 2.0148029605921187e-06, "loss": 0.0035, "step": 19973 }, { "epoch": 6.2, "learning_rate": 2.0144028805761154e-06, "loss": 0.0011, "step": 19974 }, { "epoch": 6.2, "learning_rate": 2.014002800560112e-06, "loss": 0.002, "step": 19975 }, { "epoch": 6.2, "learning_rate": 2.0136027205441088e-06, "loss": 0.0011, "step": 19976 }, { "epoch": 6.2, "learning_rate": 2.013202640528106e-06, "loss": 0.0198, "step": 19977 }, { "epoch": 6.2, "learning_rate": 2.0128025605121026e-06, "loss": 0.0097, "step": 19978 }, { "epoch": 6.2, "learning_rate": 2.0124024804960992e-06, "loss": 0.0326, "step": 19979 }, { "epoch": 6.2, "learning_rate": 2.0120024004800964e-06, "loss": 0.0008, "step": 19980 }, { "epoch": 6.2, "learning_rate": 2.011602320464093e-06, "loss": 0.0059, "step": 19981 }, { "epoch": 6.2, "learning_rate": 2.0112022404480897e-06, "loss": 0.0017, "step": 19982 }, { "epoch": 6.2, "learning_rate": 2.0108021604320864e-06, "loss": 0.0008, "step": 19983 }, { "epoch": 6.2, "learning_rate": 2.0104020804160835e-06, "loss": 0.0023, "step": 19984 }, { "epoch": 6.2, "learning_rate": 2.01000200040008e-06, "loss": 0.0011, "step": 19985 }, { "epoch": 6.2, "learning_rate": 2.009601920384077e-06, "loss": 0.0014, "step": 19986 }, { "epoch": 6.2, "learning_rate": 2.009201840368074e-06, "loss": 0.0078, "step": 19987 }, { "epoch": 6.2, "learning_rate": 2.0088017603520707e-06, "loss": 0.0125, "step": 19988 }, { "epoch": 6.2, "learning_rate": 2.0084016803360674e-06, "loss": 0.0031, "step": 19989 }, { "epoch": 6.2, "learning_rate": 2.008001600320064e-06, "loss": 0.0007, "step": 19990 }, { "epoch": 6.2, "learning_rate": 2.007601520304061e-06, "loss": 0.0288, "step": 19991 }, { "epoch": 6.21, "learning_rate": 2.007201440288058e-06, "loss": 0.0022, "step": 19992 }, { "epoch": 6.21, "learning_rate": 2.0068013602720545e-06, "loss": 0.0018, "step": 19993 }, { "epoch": 6.21, "learning_rate": 2.0064012802560517e-06, "loss": 0.0017, "step": 19994 }, { "epoch": 6.21, "learning_rate": 2.006001200240048e-06, "loss": 0.0023, "step": 19995 }, { "epoch": 6.21, "learning_rate": 2.005601120224045e-06, "loss": 0.0018, "step": 19996 }, { "epoch": 6.21, "learning_rate": 2.0052010402080417e-06, "loss": 0.0018, "step": 19997 }, { "epoch": 6.21, "learning_rate": 2.0048009601920384e-06, "loss": 0.0195, "step": 19998 }, { "epoch": 6.21, "learning_rate": 2.0044008801760355e-06, "loss": 0.0038, "step": 19999 }, { "epoch": 6.21, "learning_rate": 2.004000800160032e-06, "loss": 0.002, "step": 20000 }, { "epoch": 6.21, "learning_rate": 2.003600720144029e-06, "loss": 0.0014, "step": 20001 }, { "epoch": 6.21, "learning_rate": 2.0032006401280256e-06, "loss": 0.0064, "step": 20002 }, { "epoch": 6.21, "learning_rate": 2.0028005601120227e-06, "loss": 0.001, "step": 20003 }, { "epoch": 6.21, "learning_rate": 2.0024004800960194e-06, "loss": 0.002, "step": 20004 }, { "epoch": 6.21, "learning_rate": 2.002000400080016e-06, "loss": 0.0066, "step": 20005 }, { "epoch": 6.21, "learning_rate": 2.001600320064013e-06, "loss": 0.0017, "step": 20006 }, { "epoch": 6.21, "learning_rate": 2.00120024004801e-06, "loss": 0.0016, "step": 20007 }, { "epoch": 6.21, "learning_rate": 2.0008001600320065e-06, "loss": 0.0033, "step": 20008 }, { "epoch": 6.21, "learning_rate": 2.0004000800160032e-06, "loss": 0.0026, "step": 20009 }, { "epoch": 6.21, "learning_rate": 2.0000000000000003e-06, "loss": 0.0033, "step": 20010 }, { "epoch": 6.21, "learning_rate": 1.999599919983997e-06, "loss": 0.0061, "step": 20011 }, { "epoch": 6.21, "learning_rate": 1.9991998399679937e-06, "loss": 0.0039, "step": 20012 }, { "epoch": 6.21, "learning_rate": 1.998799759951991e-06, "loss": 0.0012, "step": 20013 }, { "epoch": 6.21, "learning_rate": 1.998399679935987e-06, "loss": 0.0024, "step": 20014 }, { "epoch": 6.21, "learning_rate": 1.997999599919984e-06, "loss": 0.0028, "step": 20015 }, { "epoch": 6.21, "learning_rate": 1.997599519903981e-06, "loss": 0.0027, "step": 20016 }, { "epoch": 6.21, "learning_rate": 1.9971994398879776e-06, "loss": 0.0027, "step": 20017 }, { "epoch": 6.21, "learning_rate": 1.9967993598719747e-06, "loss": 0.0082, "step": 20018 }, { "epoch": 6.21, "learning_rate": 1.9963992798559714e-06, "loss": 0.0008, "step": 20019 }, { "epoch": 6.21, "learning_rate": 1.995999199839968e-06, "loss": 0.0013, "step": 20020 }, { "epoch": 6.21, "learning_rate": 1.9955991198239647e-06, "loss": 0.0012, "step": 20021 }, { "epoch": 6.21, "learning_rate": 1.995199039807962e-06, "loss": 0.0024, "step": 20022 }, { "epoch": 6.21, "learning_rate": 1.9947989597919585e-06, "loss": 0.0015, "step": 20023 }, { "epoch": 6.22, "learning_rate": 1.994398879775955e-06, "loss": 0.0628, "step": 20024 }, { "epoch": 6.22, "learning_rate": 1.9939987997599523e-06, "loss": 0.0014, "step": 20025 }, { "epoch": 6.22, "learning_rate": 1.993598719743949e-06, "loss": 0.002, "step": 20026 }, { "epoch": 6.22, "learning_rate": 1.9931986397279457e-06, "loss": 0.0057, "step": 20027 }, { "epoch": 6.22, "learning_rate": 1.9927985597119424e-06, "loss": 0.0081, "step": 20028 }, { "epoch": 6.22, "learning_rate": 1.9923984796959395e-06, "loss": 0.0034, "step": 20029 }, { "epoch": 6.22, "learning_rate": 1.991998399679936e-06, "loss": 0.002, "step": 20030 }, { "epoch": 6.22, "learning_rate": 1.991598319663933e-06, "loss": 0.0028, "step": 20031 }, { "epoch": 6.22, "learning_rate": 1.99119823964793e-06, "loss": 0.0062, "step": 20032 }, { "epoch": 6.22, "learning_rate": 1.9907981596319267e-06, "loss": 0.001, "step": 20033 }, { "epoch": 6.22, "learning_rate": 1.9903980796159233e-06, "loss": 0.0058, "step": 20034 }, { "epoch": 6.22, "learning_rate": 1.98999799959992e-06, "loss": 0.001, "step": 20035 }, { "epoch": 6.22, "learning_rate": 1.9895979195839167e-06, "loss": 0.015, "step": 20036 }, { "epoch": 6.22, "learning_rate": 1.989197839567914e-06, "loss": 0.0018, "step": 20037 }, { "epoch": 6.22, "learning_rate": 1.9887977595519105e-06, "loss": 0.0157, "step": 20038 }, { "epoch": 6.22, "learning_rate": 1.988397679535907e-06, "loss": 0.0026, "step": 20039 }, { "epoch": 6.22, "learning_rate": 1.9879975995199043e-06, "loss": 0.0042, "step": 20040 }, { "epoch": 6.22, "learning_rate": 1.987597519503901e-06, "loss": 0.0055, "step": 20041 }, { "epoch": 6.22, "learning_rate": 1.9871974394878977e-06, "loss": 0.0011, "step": 20042 }, { "epoch": 6.22, "learning_rate": 1.9867973594718944e-06, "loss": 0.001, "step": 20043 }, { "epoch": 6.22, "learning_rate": 1.9863972794558915e-06, "loss": 0.0021, "step": 20044 }, { "epoch": 6.22, "learning_rate": 1.985997199439888e-06, "loss": 0.0009, "step": 20045 }, { "epoch": 6.22, "learning_rate": 1.985597119423885e-06, "loss": 0.0009, "step": 20046 }, { "epoch": 6.22, "learning_rate": 1.985197039407882e-06, "loss": 0.0028, "step": 20047 }, { "epoch": 6.22, "learning_rate": 1.9847969593918786e-06, "loss": 0.0027, "step": 20048 }, { "epoch": 6.22, "learning_rate": 1.9843968793758753e-06, "loss": 0.0018, "step": 20049 }, { "epoch": 6.22, "learning_rate": 1.983996799359872e-06, "loss": 0.0007, "step": 20050 }, { "epoch": 6.22, "learning_rate": 1.983596719343869e-06, "loss": 0.0019, "step": 20051 }, { "epoch": 6.22, "learning_rate": 1.983196639327866e-06, "loss": 0.0013, "step": 20052 }, { "epoch": 6.22, "learning_rate": 1.9827965593118625e-06, "loss": 0.0014, "step": 20053 }, { "epoch": 6.22, "learning_rate": 1.9823964792958596e-06, "loss": 0.0029, "step": 20054 }, { "epoch": 6.22, "learning_rate": 1.981996399279856e-06, "loss": 0.0021, "step": 20055 }, { "epoch": 6.22, "learning_rate": 1.981596319263853e-06, "loss": 0.0013, "step": 20056 }, { "epoch": 6.23, "learning_rate": 1.9811962392478497e-06, "loss": 0.0065, "step": 20057 }, { "epoch": 6.23, "learning_rate": 1.9807961592318464e-06, "loss": 0.0069, "step": 20058 }, { "epoch": 6.23, "learning_rate": 1.9803960792158435e-06, "loss": 0.0049, "step": 20059 }, { "epoch": 6.23, "learning_rate": 1.97999599919984e-06, "loss": 0.0021, "step": 20060 }, { "epoch": 6.23, "learning_rate": 1.979595919183837e-06, "loss": 0.0052, "step": 20061 }, { "epoch": 6.23, "learning_rate": 1.9791958391678335e-06, "loss": 0.0023, "step": 20062 }, { "epoch": 6.23, "learning_rate": 1.9787957591518306e-06, "loss": 0.0013, "step": 20063 }, { "epoch": 6.23, "learning_rate": 1.9783956791358273e-06, "loss": 0.001, "step": 20064 }, { "epoch": 6.23, "learning_rate": 1.977995599119824e-06, "loss": 0.0052, "step": 20065 }, { "epoch": 6.23, "learning_rate": 1.977595519103821e-06, "loss": 0.0069, "step": 20066 }, { "epoch": 6.23, "learning_rate": 1.977195439087818e-06, "loss": 0.0053, "step": 20067 }, { "epoch": 6.23, "learning_rate": 1.9767953590718145e-06, "loss": 0.0035, "step": 20068 }, { "epoch": 6.23, "learning_rate": 1.976395279055811e-06, "loss": 0.0042, "step": 20069 }, { "epoch": 6.23, "learning_rate": 1.9759951990398083e-06, "loss": 0.0015, "step": 20070 }, { "epoch": 6.23, "learning_rate": 1.975595119023805e-06, "loss": 0.0032, "step": 20071 }, { "epoch": 6.23, "learning_rate": 1.9751950390078017e-06, "loss": 0.0019, "step": 20072 }, { "epoch": 6.23, "learning_rate": 1.9747949589917988e-06, "loss": 0.0056, "step": 20073 }, { "epoch": 6.23, "learning_rate": 1.974394878975795e-06, "loss": 0.0021, "step": 20074 }, { "epoch": 6.23, "learning_rate": 1.973994798959792e-06, "loss": 0.0526, "step": 20075 }, { "epoch": 6.23, "learning_rate": 1.973594718943789e-06, "loss": 0.0008, "step": 20076 }, { "epoch": 6.23, "learning_rate": 1.9731946389277855e-06, "loss": 0.0038, "step": 20077 }, { "epoch": 6.23, "learning_rate": 1.9727945589117826e-06, "loss": 0.003, "step": 20078 }, { "epoch": 6.23, "learning_rate": 1.9723944788957793e-06, "loss": 0.0023, "step": 20079 }, { "epoch": 6.23, "learning_rate": 1.971994398879776e-06, "loss": 0.0018, "step": 20080 }, { "epoch": 6.23, "learning_rate": 1.9715943188637727e-06, "loss": 0.0035, "step": 20081 }, { "epoch": 6.23, "learning_rate": 1.97119423884777e-06, "loss": 0.0017, "step": 20082 }, { "epoch": 6.23, "learning_rate": 1.9707941588317665e-06, "loss": 0.0023, "step": 20083 }, { "epoch": 6.23, "learning_rate": 1.970394078815763e-06, "loss": 0.0053, "step": 20084 }, { "epoch": 6.23, "learning_rate": 1.9699939987997603e-06, "loss": 0.001, "step": 20085 }, { "epoch": 6.23, "learning_rate": 1.969593918783757e-06, "loss": 0.0032, "step": 20086 }, { "epoch": 6.23, "learning_rate": 1.9691938387677536e-06, "loss": 0.0336, "step": 20087 }, { "epoch": 6.23, "learning_rate": 1.9687937587517508e-06, "loss": 0.0039, "step": 20088 }, { "epoch": 6.24, "learning_rate": 1.9683936787357474e-06, "loss": 0.0047, "step": 20089 }, { "epoch": 6.24, "learning_rate": 1.967993598719744e-06, "loss": 0.0038, "step": 20090 }, { "epoch": 6.24, "learning_rate": 1.967593518703741e-06, "loss": 0.0012, "step": 20091 }, { "epoch": 6.24, "learning_rate": 1.967193438687738e-06, "loss": 0.0023, "step": 20092 }, { "epoch": 6.24, "learning_rate": 1.9667933586717346e-06, "loss": 0.001, "step": 20093 }, { "epoch": 6.24, "learning_rate": 1.9663932786557313e-06, "loss": 0.0303, "step": 20094 }, { "epoch": 6.24, "learning_rate": 1.9659931986397284e-06, "loss": 0.002, "step": 20095 }, { "epoch": 6.24, "learning_rate": 1.9655931186237247e-06, "loss": 0.0089, "step": 20096 }, { "epoch": 6.24, "learning_rate": 1.9651930386077218e-06, "loss": 0.0053, "step": 20097 }, { "epoch": 6.24, "learning_rate": 1.9647929585917185e-06, "loss": 0.0047, "step": 20098 }, { "epoch": 6.24, "learning_rate": 1.964392878575715e-06, "loss": 0.0029, "step": 20099 }, { "epoch": 6.24, "learning_rate": 1.9639927985597123e-06, "loss": 0.0017, "step": 20100 }, { "epoch": 6.24, "learning_rate": 1.963592718543709e-06, "loss": 0.0041, "step": 20101 }, { "epoch": 6.24, "learning_rate": 1.9631926385277056e-06, "loss": 0.004, "step": 20102 }, { "epoch": 6.24, "learning_rate": 1.9627925585117023e-06, "loss": 0.0037, "step": 20103 }, { "epoch": 6.24, "learning_rate": 1.9623924784956994e-06, "loss": 0.0015, "step": 20104 }, { "epoch": 6.24, "learning_rate": 1.961992398479696e-06, "loss": 0.004, "step": 20105 }, { "epoch": 6.24, "learning_rate": 1.961592318463693e-06, "loss": 0.0059, "step": 20106 }, { "epoch": 6.24, "learning_rate": 1.96119223844769e-06, "loss": 0.0043, "step": 20107 }, { "epoch": 6.24, "learning_rate": 1.9607921584316866e-06, "loss": 0.0301, "step": 20108 }, { "epoch": 6.24, "learning_rate": 1.9603920784156833e-06, "loss": 0.0031, "step": 20109 }, { "epoch": 6.24, "learning_rate": 1.95999199839968e-06, "loss": 0.0016, "step": 20110 }, { "epoch": 6.24, "learning_rate": 1.959591918383677e-06, "loss": 0.0041, "step": 20111 }, { "epoch": 6.24, "learning_rate": 1.9591918383676738e-06, "loss": 0.002, "step": 20112 }, { "epoch": 6.24, "learning_rate": 1.9587917583516705e-06, "loss": 0.002, "step": 20113 }, { "epoch": 6.24, "learning_rate": 1.9583916783356676e-06, "loss": 0.0016, "step": 20114 }, { "epoch": 6.24, "learning_rate": 1.957991598319664e-06, "loss": 0.0026, "step": 20115 }, { "epoch": 6.24, "learning_rate": 1.957591518303661e-06, "loss": 0.0013, "step": 20116 }, { "epoch": 6.24, "learning_rate": 1.9571914382876576e-06, "loss": 0.0011, "step": 20117 }, { "epoch": 6.24, "learning_rate": 1.9567913582716543e-06, "loss": 0.0021, "step": 20118 }, { "epoch": 6.24, "learning_rate": 1.9563912782556514e-06, "loss": 0.0011, "step": 20119 }, { "epoch": 6.24, "learning_rate": 1.955991198239648e-06, "loss": 0.0012, "step": 20120 }, { "epoch": 6.25, "learning_rate": 1.955591118223645e-06, "loss": 0.0051, "step": 20121 }, { "epoch": 6.25, "learning_rate": 1.9551910382076415e-06, "loss": 0.028, "step": 20122 }, { "epoch": 6.25, "learning_rate": 1.9547909581916386e-06, "loss": 0.0023, "step": 20123 }, { "epoch": 6.25, "learning_rate": 1.9543908781756353e-06, "loss": 0.0048, "step": 20124 }, { "epoch": 6.25, "learning_rate": 1.953990798159632e-06, "loss": 0.0017, "step": 20125 }, { "epoch": 6.25, "learning_rate": 1.953590718143629e-06, "loss": 0.0259, "step": 20126 }, { "epoch": 6.25, "learning_rate": 1.9531906381276258e-06, "loss": 0.0562, "step": 20127 }, { "epoch": 6.25, "learning_rate": 1.9527905581116224e-06, "loss": 0.0011, "step": 20128 }, { "epoch": 6.25, "learning_rate": 1.952390478095619e-06, "loss": 0.0029, "step": 20129 }, { "epoch": 6.25, "learning_rate": 1.9519903980796162e-06, "loss": 0.0009, "step": 20130 }, { "epoch": 6.25, "learning_rate": 1.951590318063613e-06, "loss": 0.0005, "step": 20131 }, { "epoch": 6.25, "learning_rate": 1.9511902380476096e-06, "loss": 0.0422, "step": 20132 }, { "epoch": 6.25, "learning_rate": 1.9507901580316067e-06, "loss": 0.0012, "step": 20133 }, { "epoch": 6.25, "learning_rate": 1.9503900780156034e-06, "loss": 0.0096, "step": 20134 }, { "epoch": 6.25, "learning_rate": 1.9499899979996e-06, "loss": 0.0008, "step": 20135 }, { "epoch": 6.25, "learning_rate": 1.9495899179835968e-06, "loss": 0.0051, "step": 20136 }, { "epoch": 6.25, "learning_rate": 1.9491898379675935e-06, "loss": 0.0048, "step": 20137 }, { "epoch": 6.25, "learning_rate": 1.9487897579515906e-06, "loss": 0.0019, "step": 20138 }, { "epoch": 6.25, "learning_rate": 1.9483896779355873e-06, "loss": 0.0046, "step": 20139 }, { "epoch": 6.25, "learning_rate": 1.947989597919584e-06, "loss": 0.0008, "step": 20140 }, { "epoch": 6.25, "learning_rate": 1.947589517903581e-06, "loss": 0.0012, "step": 20141 }, { "epoch": 6.25, "learning_rate": 1.9471894378875777e-06, "loss": 0.0024, "step": 20142 }, { "epoch": 6.25, "learning_rate": 1.9467893578715744e-06, "loss": 0.0241, "step": 20143 }, { "epoch": 6.25, "learning_rate": 1.946389277855571e-06, "loss": 0.0013, "step": 20144 }, { "epoch": 6.25, "learning_rate": 1.9459891978395682e-06, "loss": 0.0014, "step": 20145 }, { "epoch": 6.25, "learning_rate": 1.945589117823565e-06, "loss": 0.0015, "step": 20146 }, { "epoch": 6.25, "learning_rate": 1.9451890378075616e-06, "loss": 0.0063, "step": 20147 }, { "epoch": 6.25, "learning_rate": 1.9447889577915587e-06, "loss": 0.001, "step": 20148 }, { "epoch": 6.25, "learning_rate": 1.9443888777755554e-06, "loss": 0.017, "step": 20149 }, { "epoch": 6.25, "learning_rate": 1.943988797759552e-06, "loss": 0.0018, "step": 20150 }, { "epoch": 6.25, "learning_rate": 1.9435887177435488e-06, "loss": 0.0013, "step": 20151 }, { "epoch": 6.25, "learning_rate": 1.943188637727546e-06, "loss": 0.0045, "step": 20152 }, { "epoch": 6.26, "learning_rate": 1.9427885577115426e-06, "loss": 0.0018, "step": 20153 }, { "epoch": 6.26, "learning_rate": 1.9423884776955393e-06, "loss": 0.0011, "step": 20154 }, { "epoch": 6.26, "learning_rate": 1.9419883976795364e-06, "loss": 0.0368, "step": 20155 }, { "epoch": 6.26, "learning_rate": 1.9415883176635326e-06, "loss": 0.0111, "step": 20156 }, { "epoch": 6.26, "learning_rate": 1.9411882376475297e-06, "loss": 0.0023, "step": 20157 }, { "epoch": 6.26, "learning_rate": 1.9407881576315264e-06, "loss": 0.0044, "step": 20158 }, { "epoch": 6.26, "learning_rate": 1.940388077615523e-06, "loss": 0.0041, "step": 20159 }, { "epoch": 6.26, "learning_rate": 1.9399879975995202e-06, "loss": 0.003, "step": 20160 }, { "epoch": 6.26, "learning_rate": 1.939587917583517e-06, "loss": 0.0011, "step": 20161 }, { "epoch": 6.26, "learning_rate": 1.9391878375675136e-06, "loss": 0.0033, "step": 20162 }, { "epoch": 6.26, "learning_rate": 1.9387877575515103e-06, "loss": 0.0087, "step": 20163 }, { "epoch": 6.26, "learning_rate": 1.9383876775355074e-06, "loss": 0.0063, "step": 20164 }, { "epoch": 6.26, "learning_rate": 1.937987597519504e-06, "loss": 0.0207, "step": 20165 }, { "epoch": 6.26, "learning_rate": 1.9375875175035008e-06, "loss": 0.0078, "step": 20166 }, { "epoch": 6.26, "learning_rate": 1.937187437487498e-06, "loss": 0.0006, "step": 20167 }, { "epoch": 6.26, "learning_rate": 1.936787357471494e-06, "loss": 0.0013, "step": 20168 }, { "epoch": 6.26, "learning_rate": 1.9363872774554912e-06, "loss": 0.0065, "step": 20169 }, { "epoch": 6.26, "learning_rate": 1.935987197439488e-06, "loss": 0.0011, "step": 20170 }, { "epoch": 6.26, "learning_rate": 1.935587117423485e-06, "loss": 0.0018, "step": 20171 }, { "epoch": 6.26, "learning_rate": 1.9351870374074817e-06, "loss": 0.0014, "step": 20172 }, { "epoch": 6.26, "learning_rate": 1.9347869573914784e-06, "loss": 0.0022, "step": 20173 }, { "epoch": 6.26, "learning_rate": 1.9343868773754755e-06, "loss": 0.0016, "step": 20174 }, { "epoch": 6.26, "learning_rate": 1.9339867973594718e-06, "loss": 0.0024, "step": 20175 }, { "epoch": 6.26, "learning_rate": 1.933586717343469e-06, "loss": 0.0029, "step": 20176 }, { "epoch": 6.26, "learning_rate": 1.9331866373274656e-06, "loss": 0.0042, "step": 20177 }, { "epoch": 6.26, "learning_rate": 1.9327865573114623e-06, "loss": 0.0023, "step": 20178 }, { "epoch": 6.26, "learning_rate": 1.9323864772954594e-06, "loss": 0.0009, "step": 20179 }, { "epoch": 6.26, "learning_rate": 1.931986397279456e-06, "loss": 0.0076, "step": 20180 }, { "epoch": 6.26, "learning_rate": 1.9315863172634527e-06, "loss": 0.0009, "step": 20181 }, { "epoch": 6.26, "learning_rate": 1.9311862372474494e-06, "loss": 0.0036, "step": 20182 }, { "epoch": 6.26, "learning_rate": 1.9307861572314465e-06, "loss": 0.0033, "step": 20183 }, { "epoch": 6.26, "learning_rate": 1.9303860772154432e-06, "loss": 0.0491, "step": 20184 }, { "epoch": 6.26, "learning_rate": 1.92998599719944e-06, "loss": 0.0013, "step": 20185 }, { "epoch": 6.27, "learning_rate": 1.929585917183437e-06, "loss": 0.0012, "step": 20186 }, { "epoch": 6.27, "learning_rate": 1.9291858371674337e-06, "loss": 0.0016, "step": 20187 }, { "epoch": 6.27, "learning_rate": 1.9287857571514304e-06, "loss": 0.0019, "step": 20188 }, { "epoch": 6.27, "learning_rate": 1.928385677135427e-06, "loss": 0.0039, "step": 20189 }, { "epoch": 6.27, "learning_rate": 1.927985597119424e-06, "loss": 0.0327, "step": 20190 }, { "epoch": 6.27, "learning_rate": 1.927585517103421e-06, "loss": 0.0139, "step": 20191 }, { "epoch": 6.27, "learning_rate": 1.9271854370874176e-06, "loss": 0.0044, "step": 20192 }, { "epoch": 6.27, "learning_rate": 1.9267853570714147e-06, "loss": 0.0021, "step": 20193 }, { "epoch": 6.27, "learning_rate": 1.9263852770554114e-06, "loss": 0.0046, "step": 20194 }, { "epoch": 6.27, "learning_rate": 1.925985197039408e-06, "loss": 0.0019, "step": 20195 }, { "epoch": 6.27, "learning_rate": 1.925585117023405e-06, "loss": 0.0063, "step": 20196 }, { "epoch": 6.27, "learning_rate": 1.9251850370074014e-06, "loss": 0.0164, "step": 20197 }, { "epoch": 6.27, "learning_rate": 1.9247849569913985e-06, "loss": 0.0011, "step": 20198 }, { "epoch": 6.27, "learning_rate": 1.9243848769753952e-06, "loss": 0.0016, "step": 20199 }, { "epoch": 6.27, "learning_rate": 1.923984796959392e-06, "loss": 0.0055, "step": 20200 }, { "epoch": 6.27, "learning_rate": 1.923584716943389e-06, "loss": 0.0026, "step": 20201 }, { "epoch": 6.27, "learning_rate": 1.9231846369273857e-06, "loss": 0.0012, "step": 20202 }, { "epoch": 6.27, "learning_rate": 1.9227845569113824e-06, "loss": 0.0015, "step": 20203 }, { "epoch": 6.27, "learning_rate": 1.922384476895379e-06, "loss": 0.0029, "step": 20204 }, { "epoch": 6.27, "learning_rate": 1.921984396879376e-06, "loss": 0.0039, "step": 20205 }, { "epoch": 6.27, "learning_rate": 1.921584316863373e-06, "loss": 0.0015, "step": 20206 }, { "epoch": 6.27, "learning_rate": 1.9211842368473696e-06, "loss": 0.0015, "step": 20207 }, { "epoch": 6.27, "learning_rate": 1.9207841568313667e-06, "loss": 0.0013, "step": 20208 }, { "epoch": 6.27, "learning_rate": 1.920384076815363e-06, "loss": 0.0036, "step": 20209 }, { "epoch": 6.27, "learning_rate": 1.91998399679936e-06, "loss": 0.0015, "step": 20210 }, { "epoch": 6.27, "learning_rate": 1.9195839167833567e-06, "loss": 0.0051, "step": 20211 }, { "epoch": 6.27, "learning_rate": 1.919183836767354e-06, "loss": 0.0033, "step": 20212 }, { "epoch": 6.27, "learning_rate": 1.9187837567513505e-06, "loss": 0.004, "step": 20213 }, { "epoch": 6.27, "learning_rate": 1.918383676735347e-06, "loss": 0.0033, "step": 20214 }, { "epoch": 6.27, "learning_rate": 1.9179835967193443e-06, "loss": 0.0048, "step": 20215 }, { "epoch": 6.27, "learning_rate": 1.9175835167033406e-06, "loss": 0.028, "step": 20216 }, { "epoch": 6.27, "learning_rate": 1.9171834366873377e-06, "loss": 0.0012, "step": 20217 }, { "epoch": 6.28, "learning_rate": 1.9167833566713344e-06, "loss": 0.0024, "step": 20218 }, { "epoch": 6.28, "learning_rate": 1.916383276655331e-06, "loss": 0.0014, "step": 20219 }, { "epoch": 6.28, "learning_rate": 1.915983196639328e-06, "loss": 0.0013, "step": 20220 }, { "epoch": 6.28, "learning_rate": 1.915583116623325e-06, "loss": 0.0023, "step": 20221 }, { "epoch": 6.28, "learning_rate": 1.9151830366073215e-06, "loss": 0.0446, "step": 20222 }, { "epoch": 6.28, "learning_rate": 1.9147829565913182e-06, "loss": 0.0046, "step": 20223 }, { "epoch": 6.28, "learning_rate": 1.9143828765753153e-06, "loss": 0.0025, "step": 20224 }, { "epoch": 6.28, "learning_rate": 1.913982796559312e-06, "loss": 0.0038, "step": 20225 }, { "epoch": 6.28, "learning_rate": 1.9135827165433087e-06, "loss": 0.0014, "step": 20226 }, { "epoch": 6.28, "learning_rate": 1.913182636527306e-06, "loss": 0.0027, "step": 20227 }, { "epoch": 6.28, "learning_rate": 1.912782556511302e-06, "loss": 0.005, "step": 20228 }, { "epoch": 6.28, "learning_rate": 1.912382476495299e-06, "loss": 0.0021, "step": 20229 }, { "epoch": 6.28, "learning_rate": 1.911982396479296e-06, "loss": 0.0015, "step": 20230 }, { "epoch": 6.28, "learning_rate": 1.911582316463293e-06, "loss": 0.0021, "step": 20231 }, { "epoch": 6.28, "learning_rate": 1.9111822364472897e-06, "loss": 0.0013, "step": 20232 }, { "epoch": 6.28, "learning_rate": 1.9107821564312864e-06, "loss": 0.0076, "step": 20233 }, { "epoch": 6.28, "learning_rate": 1.9103820764152835e-06, "loss": 0.0014, "step": 20234 }, { "epoch": 6.28, "learning_rate": 1.90998199639928e-06, "loss": 0.0302, "step": 20235 }, { "epoch": 6.28, "learning_rate": 1.909581916383277e-06, "loss": 0.0025, "step": 20236 }, { "epoch": 6.28, "learning_rate": 1.9091818363672735e-06, "loss": 0.0014, "step": 20237 }, { "epoch": 6.28, "learning_rate": 1.9087817563512702e-06, "loss": 0.0016, "step": 20238 }, { "epoch": 6.28, "learning_rate": 1.9083816763352673e-06, "loss": 0.0231, "step": 20239 }, { "epoch": 6.28, "learning_rate": 1.907981596319264e-06, "loss": 0.0017, "step": 20240 }, { "epoch": 6.28, "learning_rate": 1.9075815163032607e-06, "loss": 0.0014, "step": 20241 }, { "epoch": 6.28, "learning_rate": 1.9071814362872576e-06, "loss": 0.0018, "step": 20242 }, { "epoch": 6.28, "learning_rate": 1.9067813562712545e-06, "loss": 0.0014, "step": 20243 }, { "epoch": 6.28, "learning_rate": 1.9063812762552512e-06, "loss": 0.0036, "step": 20244 }, { "epoch": 6.28, "learning_rate": 1.905981196239248e-06, "loss": 0.0275, "step": 20245 }, { "epoch": 6.28, "learning_rate": 1.905581116223245e-06, "loss": 0.0031, "step": 20246 }, { "epoch": 6.28, "learning_rate": 1.9051810362072415e-06, "loss": 0.0039, "step": 20247 }, { "epoch": 6.28, "learning_rate": 1.9047809561912383e-06, "loss": 0.0052, "step": 20248 }, { "epoch": 6.28, "learning_rate": 1.9043808761752352e-06, "loss": 0.0015, "step": 20249 }, { "epoch": 6.29, "learning_rate": 1.903980796159232e-06, "loss": 0.0042, "step": 20250 }, { "epoch": 6.29, "learning_rate": 1.9035807161432288e-06, "loss": 0.0007, "step": 20251 }, { "epoch": 6.29, "learning_rate": 1.9031806361272257e-06, "loss": 0.0075, "step": 20252 }, { "epoch": 6.29, "learning_rate": 1.9027805561112226e-06, "loss": 0.0011, "step": 20253 }, { "epoch": 6.29, "learning_rate": 1.902380476095219e-06, "loss": 0.0083, "step": 20254 }, { "epoch": 6.29, "learning_rate": 1.901980396079216e-06, "loss": 0.002, "step": 20255 }, { "epoch": 6.29, "learning_rate": 1.901580316063213e-06, "loss": 0.0011, "step": 20256 }, { "epoch": 6.29, "learning_rate": 1.9011802360472096e-06, "loss": 0.002, "step": 20257 }, { "epoch": 6.29, "learning_rate": 1.9007801560312065e-06, "loss": 0.0015, "step": 20258 }, { "epoch": 6.29, "learning_rate": 1.9003800760152034e-06, "loss": 0.0027, "step": 20259 }, { "epoch": 6.29, "learning_rate": 1.8999799959991999e-06, "loss": 0.0022, "step": 20260 }, { "epoch": 6.29, "learning_rate": 1.8995799159831968e-06, "loss": 0.0015, "step": 20261 }, { "epoch": 6.29, "learning_rate": 1.8991798359671937e-06, "loss": 0.0016, "step": 20262 }, { "epoch": 6.29, "learning_rate": 1.8987797559511903e-06, "loss": 0.0075, "step": 20263 }, { "epoch": 6.29, "learning_rate": 1.8983796759351872e-06, "loss": 0.0011, "step": 20264 }, { "epoch": 6.29, "learning_rate": 1.8979795959191841e-06, "loss": 0.0084, "step": 20265 }, { "epoch": 6.29, "learning_rate": 1.8975795159031806e-06, "loss": 0.001, "step": 20266 }, { "epoch": 6.29, "learning_rate": 1.8971794358871775e-06, "loss": 0.0012, "step": 20267 }, { "epoch": 6.29, "learning_rate": 1.8967793558711744e-06, "loss": 0.0012, "step": 20268 }, { "epoch": 6.29, "learning_rate": 1.896379275855171e-06, "loss": 0.0025, "step": 20269 }, { "epoch": 6.29, "learning_rate": 1.895979195839168e-06, "loss": 0.006, "step": 20270 }, { "epoch": 6.29, "learning_rate": 1.8955791158231649e-06, "loss": 0.0018, "step": 20271 }, { "epoch": 6.29, "learning_rate": 1.8951790358071614e-06, "loss": 0.0093, "step": 20272 }, { "epoch": 6.29, "learning_rate": 1.8947789557911583e-06, "loss": 0.0011, "step": 20273 }, { "epoch": 6.29, "learning_rate": 1.8943788757751552e-06, "loss": 0.0015, "step": 20274 }, { "epoch": 6.29, "learning_rate": 1.893978795759152e-06, "loss": 0.0017, "step": 20275 }, { "epoch": 6.29, "learning_rate": 1.8935787157431487e-06, "loss": 0.0024, "step": 20276 }, { "epoch": 6.29, "learning_rate": 1.8931786357271456e-06, "loss": 0.0027, "step": 20277 }, { "epoch": 6.29, "learning_rate": 1.8927785557111425e-06, "loss": 0.0017, "step": 20278 }, { "epoch": 6.29, "learning_rate": 1.892378475695139e-06, "loss": 0.0019, "step": 20279 }, { "epoch": 6.29, "learning_rate": 1.891978395679136e-06, "loss": 0.0014, "step": 20280 }, { "epoch": 6.29, "learning_rate": 1.8915783156631328e-06, "loss": 0.0204, "step": 20281 }, { "epoch": 6.3, "learning_rate": 1.8911782356471295e-06, "loss": 0.0019, "step": 20282 }, { "epoch": 6.3, "learning_rate": 1.8907781556311264e-06, "loss": 0.005, "step": 20283 }, { "epoch": 6.3, "learning_rate": 1.8903780756151233e-06, "loss": 0.001, "step": 20284 }, { "epoch": 6.3, "learning_rate": 1.88997799559912e-06, "loss": 0.0279, "step": 20285 }, { "epoch": 6.3, "learning_rate": 1.8895779155831167e-06, "loss": 0.0014, "step": 20286 }, { "epoch": 6.3, "learning_rate": 1.8891778355671136e-06, "loss": 0.0019, "step": 20287 }, { "epoch": 6.3, "learning_rate": 1.8887777555511102e-06, "loss": 0.0018, "step": 20288 }, { "epoch": 6.3, "learning_rate": 1.8883776755351071e-06, "loss": 0.0021, "step": 20289 }, { "epoch": 6.3, "learning_rate": 1.887977595519104e-06, "loss": 0.0513, "step": 20290 }, { "epoch": 6.3, "learning_rate": 1.8875775155031007e-06, "loss": 0.0045, "step": 20291 }, { "epoch": 6.3, "learning_rate": 1.8871774354870976e-06, "loss": 0.0034, "step": 20292 }, { "epoch": 6.3, "learning_rate": 1.8867773554710943e-06, "loss": 0.0034, "step": 20293 }, { "epoch": 6.3, "learning_rate": 1.8863772754550912e-06, "loss": 0.0028, "step": 20294 }, { "epoch": 6.3, "learning_rate": 1.885977195439088e-06, "loss": 0.003, "step": 20295 }, { "epoch": 6.3, "learning_rate": 1.8855771154230848e-06, "loss": 0.0013, "step": 20296 }, { "epoch": 6.3, "learning_rate": 1.8851770354070817e-06, "loss": 0.0026, "step": 20297 }, { "epoch": 6.3, "learning_rate": 1.8847769553910784e-06, "loss": 0.0023, "step": 20298 }, { "epoch": 6.3, "learning_rate": 1.8843768753750753e-06, "loss": 0.0023, "step": 20299 }, { "epoch": 6.3, "learning_rate": 1.8839767953590722e-06, "loss": 0.003, "step": 20300 }, { "epoch": 6.3, "learning_rate": 1.8835767153430687e-06, "loss": 0.0029, "step": 20301 }, { "epoch": 6.3, "learning_rate": 1.8831766353270655e-06, "loss": 0.0006, "step": 20302 }, { "epoch": 6.3, "learning_rate": 1.8827765553110624e-06, "loss": 0.0038, "step": 20303 }, { "epoch": 6.3, "learning_rate": 1.8823764752950591e-06, "loss": 0.0016, "step": 20304 }, { "epoch": 6.3, "learning_rate": 1.881976395279056e-06, "loss": 0.0018, "step": 20305 }, { "epoch": 6.3, "learning_rate": 1.881576315263053e-06, "loss": 0.001, "step": 20306 }, { "epoch": 6.3, "learning_rate": 1.8811762352470494e-06, "loss": 0.0029, "step": 20307 }, { "epoch": 6.3, "learning_rate": 1.8807761552310463e-06, "loss": 0.0018, "step": 20308 }, { "epoch": 6.3, "learning_rate": 1.8803760752150432e-06, "loss": 0.0018, "step": 20309 }, { "epoch": 6.3, "learning_rate": 1.8799759951990399e-06, "loss": 0.0016, "step": 20310 }, { "epoch": 6.3, "learning_rate": 1.8795759151830368e-06, "loss": 0.0007, "step": 20311 }, { "epoch": 6.3, "learning_rate": 1.8791758351670337e-06, "loss": 0.0037, "step": 20312 }, { "epoch": 6.3, "learning_rate": 1.8787757551510302e-06, "loss": 0.0097, "step": 20313 }, { "epoch": 6.31, "learning_rate": 1.878375675135027e-06, "loss": 0.0013, "step": 20314 }, { "epoch": 6.31, "learning_rate": 1.877975595119024e-06, "loss": 0.0019, "step": 20315 }, { "epoch": 6.31, "learning_rate": 1.8775755151030209e-06, "loss": 0.0009, "step": 20316 }, { "epoch": 6.31, "learning_rate": 1.8771754350870175e-06, "loss": 0.0012, "step": 20317 }, { "epoch": 6.31, "learning_rate": 1.8767753550710144e-06, "loss": 0.0034, "step": 20318 }, { "epoch": 6.31, "learning_rate": 1.8763752750550113e-06, "loss": 0.0036, "step": 20319 }, { "epoch": 6.31, "learning_rate": 1.8759751950390078e-06, "loss": 0.0039, "step": 20320 }, { "epoch": 6.31, "learning_rate": 1.8755751150230047e-06, "loss": 0.0024, "step": 20321 }, { "epoch": 6.31, "learning_rate": 1.8751750350070016e-06, "loss": 0.0026, "step": 20322 }, { "epoch": 6.31, "learning_rate": 1.8747749549909983e-06, "loss": 0.0038, "step": 20323 }, { "epoch": 6.31, "learning_rate": 1.8743748749749952e-06, "loss": 0.0416, "step": 20324 }, { "epoch": 6.31, "learning_rate": 1.873974794958992e-06, "loss": 0.0036, "step": 20325 }, { "epoch": 6.31, "learning_rate": 1.8735747149429886e-06, "loss": 0.0013, "step": 20326 }, { "epoch": 6.31, "learning_rate": 1.8731746349269855e-06, "loss": 0.0049, "step": 20327 }, { "epoch": 6.31, "learning_rate": 1.8727745549109824e-06, "loss": 0.0017, "step": 20328 }, { "epoch": 6.31, "learning_rate": 1.872374474894979e-06, "loss": 0.0041, "step": 20329 }, { "epoch": 6.31, "learning_rate": 1.871974394878976e-06, "loss": 0.002, "step": 20330 }, { "epoch": 6.31, "learning_rate": 1.8715743148629728e-06, "loss": 0.0031, "step": 20331 }, { "epoch": 6.31, "learning_rate": 1.8711742348469693e-06, "loss": 0.0015, "step": 20332 }, { "epoch": 6.31, "learning_rate": 1.8707741548309662e-06, "loss": 0.0008, "step": 20333 }, { "epoch": 6.31, "learning_rate": 1.8703740748149631e-06, "loss": 0.0029, "step": 20334 }, { "epoch": 6.31, "learning_rate": 1.86997399479896e-06, "loss": 0.0262, "step": 20335 }, { "epoch": 6.31, "learning_rate": 1.8695739147829567e-06, "loss": 0.0023, "step": 20336 }, { "epoch": 6.31, "learning_rate": 1.8691738347669536e-06, "loss": 0.0148, "step": 20337 }, { "epoch": 6.31, "learning_rate": 1.8687737547509505e-06, "loss": 0.0018, "step": 20338 }, { "epoch": 6.31, "learning_rate": 1.8683736747349472e-06, "loss": 0.0009, "step": 20339 }, { "epoch": 6.31, "learning_rate": 1.8679735947189439e-06, "loss": 0.0038, "step": 20340 }, { "epoch": 6.31, "learning_rate": 1.8675735147029408e-06, "loss": 0.0411, "step": 20341 }, { "epoch": 6.31, "learning_rate": 1.8671734346869374e-06, "loss": 0.001, "step": 20342 }, { "epoch": 6.31, "learning_rate": 1.8667733546709343e-06, "loss": 0.0048, "step": 20343 }, { "epoch": 6.31, "learning_rate": 1.8663732746549312e-06, "loss": 0.0005, "step": 20344 }, { "epoch": 6.31, "learning_rate": 1.865973194638928e-06, "loss": 0.0019, "step": 20345 }, { "epoch": 6.31, "learning_rate": 1.8655731146229248e-06, "loss": 0.0013, "step": 20346 }, { "epoch": 6.32, "learning_rate": 1.8651730346069215e-06, "loss": 0.0032, "step": 20347 }, { "epoch": 6.32, "learning_rate": 1.8647729545909182e-06, "loss": 0.0014, "step": 20348 }, { "epoch": 6.32, "learning_rate": 1.864372874574915e-06, "loss": 0.0014, "step": 20349 }, { "epoch": 6.32, "learning_rate": 1.863972794558912e-06, "loss": 0.0043, "step": 20350 }, { "epoch": 6.32, "learning_rate": 1.8635727145429087e-06, "loss": 0.003, "step": 20351 }, { "epoch": 6.32, "learning_rate": 1.8631726345269056e-06, "loss": 0.0009, "step": 20352 }, { "epoch": 6.32, "learning_rate": 1.8627725545109025e-06, "loss": 0.0019, "step": 20353 }, { "epoch": 6.32, "learning_rate": 1.862372474494899e-06, "loss": 0.0013, "step": 20354 }, { "epoch": 6.32, "learning_rate": 1.8619723944788959e-06, "loss": 0.003, "step": 20355 }, { "epoch": 6.32, "learning_rate": 1.8615723144628927e-06, "loss": 0.0055, "step": 20356 }, { "epoch": 6.32, "learning_rate": 1.8611722344468896e-06, "loss": 0.0082, "step": 20357 }, { "epoch": 6.32, "learning_rate": 1.8607721544308863e-06, "loss": 0.0051, "step": 20358 }, { "epoch": 6.32, "learning_rate": 1.8603720744148832e-06, "loss": 0.0226, "step": 20359 }, { "epoch": 6.32, "learning_rate": 1.8599719943988801e-06, "loss": 0.0006, "step": 20360 }, { "epoch": 6.32, "learning_rate": 1.8595719143828766e-06, "loss": 0.0022, "step": 20361 }, { "epoch": 6.32, "learning_rate": 1.8591718343668735e-06, "loss": 0.0015, "step": 20362 }, { "epoch": 6.32, "learning_rate": 1.8587717543508704e-06, "loss": 0.0042, "step": 20363 }, { "epoch": 6.32, "learning_rate": 1.858371674334867e-06, "loss": 0.0019, "step": 20364 }, { "epoch": 6.32, "learning_rate": 1.857971594318864e-06, "loss": 0.0016, "step": 20365 }, { "epoch": 6.32, "learning_rate": 1.8575715143028609e-06, "loss": 0.0011, "step": 20366 }, { "epoch": 6.32, "learning_rate": 1.8571714342868574e-06, "loss": 0.0016, "step": 20367 }, { "epoch": 6.32, "learning_rate": 1.8567713542708543e-06, "loss": 0.0049, "step": 20368 }, { "epoch": 6.32, "learning_rate": 1.8563712742548512e-06, "loss": 0.0021, "step": 20369 }, { "epoch": 6.32, "learning_rate": 1.8559711942388478e-06, "loss": 0.033, "step": 20370 }, { "epoch": 6.32, "learning_rate": 1.8555711142228447e-06, "loss": 0.0013, "step": 20371 }, { "epoch": 6.32, "learning_rate": 1.8551710342068416e-06, "loss": 0.0031, "step": 20372 }, { "epoch": 6.32, "learning_rate": 1.8547709541908381e-06, "loss": 0.0029, "step": 20373 }, { "epoch": 6.32, "learning_rate": 1.854370874174835e-06, "loss": 0.003, "step": 20374 }, { "epoch": 6.32, "learning_rate": 1.853970794158832e-06, "loss": 0.0034, "step": 20375 }, { "epoch": 6.32, "learning_rate": 1.8535707141428288e-06, "loss": 0.0013, "step": 20376 }, { "epoch": 6.32, "learning_rate": 1.8531706341268255e-06, "loss": 0.0014, "step": 20377 }, { "epoch": 6.32, "learning_rate": 1.8527705541108224e-06, "loss": 0.0019, "step": 20378 }, { "epoch": 6.33, "learning_rate": 1.8523704740948193e-06, "loss": 0.0236, "step": 20379 }, { "epoch": 6.33, "learning_rate": 1.8519703940788158e-06, "loss": 0.0014, "step": 20380 }, { "epoch": 6.33, "learning_rate": 1.8515703140628127e-06, "loss": 0.002, "step": 20381 }, { "epoch": 6.33, "learning_rate": 1.8511702340468096e-06, "loss": 0.0128, "step": 20382 }, { "epoch": 6.33, "learning_rate": 1.8507701540308062e-06, "loss": 0.0015, "step": 20383 }, { "epoch": 6.33, "learning_rate": 1.8503700740148031e-06, "loss": 0.0026, "step": 20384 }, { "epoch": 6.33, "learning_rate": 1.8499699939988e-06, "loss": 0.0156, "step": 20385 }, { "epoch": 6.33, "learning_rate": 1.8495699139827965e-06, "loss": 0.0083, "step": 20386 }, { "epoch": 6.33, "learning_rate": 1.8491698339667934e-06, "loss": 0.0031, "step": 20387 }, { "epoch": 6.33, "learning_rate": 1.8487697539507903e-06, "loss": 0.0012, "step": 20388 }, { "epoch": 6.33, "learning_rate": 1.848369673934787e-06, "loss": 0.0011, "step": 20389 }, { "epoch": 6.33, "learning_rate": 1.8479695939187839e-06, "loss": 0.0013, "step": 20390 }, { "epoch": 6.33, "learning_rate": 1.8475695139027808e-06, "loss": 0.0011, "step": 20391 }, { "epoch": 6.33, "learning_rate": 1.8471694338867775e-06, "loss": 0.0017, "step": 20392 }, { "epoch": 6.33, "learning_rate": 1.8467693538707744e-06, "loss": 0.0032, "step": 20393 }, { "epoch": 6.33, "learning_rate": 1.846369273854771e-06, "loss": 0.0112, "step": 20394 }, { "epoch": 6.33, "learning_rate": 1.8459691938387677e-06, "loss": 0.0034, "step": 20395 }, { "epoch": 6.33, "learning_rate": 1.8455691138227646e-06, "loss": 0.0037, "step": 20396 }, { "epoch": 6.33, "learning_rate": 1.8451690338067615e-06, "loss": 0.0021, "step": 20397 }, { "epoch": 6.33, "learning_rate": 1.8447689537907584e-06, "loss": 0.0042, "step": 20398 }, { "epoch": 6.33, "learning_rate": 1.8443688737747551e-06, "loss": 0.0033, "step": 20399 }, { "epoch": 6.33, "learning_rate": 1.843968793758752e-06, "loss": 0.0015, "step": 20400 }, { "epoch": 6.33, "learning_rate": 1.8435687137427487e-06, "loss": 0.0027, "step": 20401 }, { "epoch": 6.33, "learning_rate": 1.8431686337267454e-06, "loss": 0.0007, "step": 20402 }, { "epoch": 6.33, "learning_rate": 1.8427685537107423e-06, "loss": 0.0031, "step": 20403 }, { "epoch": 6.33, "learning_rate": 1.8423684736947392e-06, "loss": 0.0031, "step": 20404 }, { "epoch": 6.33, "learning_rate": 1.8419683936787359e-06, "loss": 0.0019, "step": 20405 }, { "epoch": 6.33, "learning_rate": 1.8415683136627328e-06, "loss": 0.0065, "step": 20406 }, { "epoch": 6.33, "learning_rate": 1.8411682336467297e-06, "loss": 0.0023, "step": 20407 }, { "epoch": 6.33, "learning_rate": 1.8407681536307262e-06, "loss": 0.0014, "step": 20408 }, { "epoch": 6.33, "learning_rate": 1.840368073614723e-06, "loss": 0.0202, "step": 20409 }, { "epoch": 6.33, "learning_rate": 1.83996799359872e-06, "loss": 0.0008, "step": 20410 }, { "epoch": 6.34, "learning_rate": 1.8395679135827166e-06, "loss": 0.0016, "step": 20411 }, { "epoch": 6.34, "learning_rate": 1.8391678335667135e-06, "loss": 0.0053, "step": 20412 }, { "epoch": 6.34, "learning_rate": 1.8387677535507104e-06, "loss": 0.0044, "step": 20413 }, { "epoch": 6.34, "learning_rate": 1.838367673534707e-06, "loss": 0.0011, "step": 20414 }, { "epoch": 6.34, "learning_rate": 1.8379675935187038e-06, "loss": 0.0015, "step": 20415 }, { "epoch": 6.34, "learning_rate": 1.8375675135027007e-06, "loss": 0.0012, "step": 20416 }, { "epoch": 6.34, "learning_rate": 1.8371674334866976e-06, "loss": 0.0032, "step": 20417 }, { "epoch": 6.34, "learning_rate": 1.8367673534706943e-06, "loss": 0.0053, "step": 20418 }, { "epoch": 6.34, "learning_rate": 1.8363672734546912e-06, "loss": 0.0125, "step": 20419 }, { "epoch": 6.34, "learning_rate": 1.835967193438688e-06, "loss": 0.0009, "step": 20420 }, { "epoch": 6.34, "learning_rate": 1.8355671134226846e-06, "loss": 0.001, "step": 20421 }, { "epoch": 6.34, "learning_rate": 1.8351670334066815e-06, "loss": 0.0009, "step": 20422 }, { "epoch": 6.34, "learning_rate": 1.8347669533906784e-06, "loss": 0.0012, "step": 20423 }, { "epoch": 6.34, "learning_rate": 1.834366873374675e-06, "loss": 0.0034, "step": 20424 }, { "epoch": 6.34, "learning_rate": 1.833966793358672e-06, "loss": 0.0012, "step": 20425 }, { "epoch": 6.34, "learning_rate": 1.8335667133426688e-06, "loss": 0.0017, "step": 20426 }, { "epoch": 6.34, "learning_rate": 1.8331666333266653e-06, "loss": 0.0018, "step": 20427 }, { "epoch": 6.34, "learning_rate": 1.8327665533106622e-06, "loss": 0.001, "step": 20428 }, { "epoch": 6.34, "learning_rate": 1.832366473294659e-06, "loss": 0.0019, "step": 20429 }, { "epoch": 6.34, "learning_rate": 1.8319663932786558e-06, "loss": 0.0012, "step": 20430 }, { "epoch": 6.34, "learning_rate": 1.8315663132626527e-06, "loss": 0.0027, "step": 20431 }, { "epoch": 6.34, "learning_rate": 1.8311662332466496e-06, "loss": 0.0017, "step": 20432 }, { "epoch": 6.34, "learning_rate": 1.830766153230646e-06, "loss": 0.0082, "step": 20433 }, { "epoch": 6.34, "learning_rate": 1.830366073214643e-06, "loss": 0.0015, "step": 20434 }, { "epoch": 6.34, "learning_rate": 1.8299659931986399e-06, "loss": 0.0014, "step": 20435 }, { "epoch": 6.34, "learning_rate": 1.8295659131826365e-06, "loss": 0.0017, "step": 20436 }, { "epoch": 6.34, "learning_rate": 1.8291658331666334e-06, "loss": 0.0039, "step": 20437 }, { "epoch": 6.34, "learning_rate": 1.8287657531506303e-06, "loss": 0.0024, "step": 20438 }, { "epoch": 6.34, "learning_rate": 1.8283656731346272e-06, "loss": 0.0009, "step": 20439 }, { "epoch": 6.34, "learning_rate": 1.8279655931186237e-06, "loss": 0.0011, "step": 20440 }, { "epoch": 6.34, "learning_rate": 1.8275655131026206e-06, "loss": 0.0061, "step": 20441 }, { "epoch": 6.34, "learning_rate": 1.8271654330866175e-06, "loss": 0.0184, "step": 20442 }, { "epoch": 6.35, "learning_rate": 1.8267653530706142e-06, "loss": 0.0047, "step": 20443 }, { "epoch": 6.35, "learning_rate": 1.826365273054611e-06, "loss": 0.0026, "step": 20444 }, { "epoch": 6.35, "learning_rate": 1.825965193038608e-06, "loss": 0.0024, "step": 20445 }, { "epoch": 6.35, "learning_rate": 1.8255651130226047e-06, "loss": 0.0017, "step": 20446 }, { "epoch": 6.35, "learning_rate": 1.8251650330066016e-06, "loss": 0.0014, "step": 20447 }, { "epoch": 6.35, "learning_rate": 1.8247649529905983e-06, "loss": 0.016, "step": 20448 }, { "epoch": 6.35, "learning_rate": 1.824364872974595e-06, "loss": 0.0018, "step": 20449 }, { "epoch": 6.35, "learning_rate": 1.8239647929585918e-06, "loss": 0.0038, "step": 20450 }, { "epoch": 6.35, "learning_rate": 1.8235647129425887e-06, "loss": 0.0011, "step": 20451 }, { "epoch": 6.35, "learning_rate": 1.8231646329265854e-06, "loss": 0.0018, "step": 20452 }, { "epoch": 6.35, "learning_rate": 1.8227645529105823e-06, "loss": 0.0093, "step": 20453 }, { "epoch": 6.35, "learning_rate": 1.8223644728945792e-06, "loss": 0.0015, "step": 20454 }, { "epoch": 6.35, "learning_rate": 1.8219643928785757e-06, "loss": 0.0068, "step": 20455 }, { "epoch": 6.35, "learning_rate": 1.8215643128625726e-06, "loss": 0.0015, "step": 20456 }, { "epoch": 6.35, "learning_rate": 1.8211642328465695e-06, "loss": 0.0056, "step": 20457 }, { "epoch": 6.35, "learning_rate": 1.8207641528305664e-06, "loss": 0.0016, "step": 20458 }, { "epoch": 6.35, "learning_rate": 1.820364072814563e-06, "loss": 0.0012, "step": 20459 }, { "epoch": 6.35, "learning_rate": 1.81996399279856e-06, "loss": 0.0018, "step": 20460 }, { "epoch": 6.35, "learning_rate": 1.8195639127825569e-06, "loss": 0.003, "step": 20461 }, { "epoch": 6.35, "learning_rate": 1.8191638327665534e-06, "loss": 0.0015, "step": 20462 }, { "epoch": 6.35, "learning_rate": 1.8187637527505502e-06, "loss": 0.0011, "step": 20463 }, { "epoch": 6.35, "learning_rate": 1.8183636727345471e-06, "loss": 0.0031, "step": 20464 }, { "epoch": 6.35, "learning_rate": 1.8179635927185438e-06, "loss": 0.0193, "step": 20465 }, { "epoch": 6.35, "learning_rate": 1.8175635127025407e-06, "loss": 0.0044, "step": 20466 }, { "epoch": 6.35, "learning_rate": 1.8171634326865376e-06, "loss": 0.0011, "step": 20467 }, { "epoch": 6.35, "learning_rate": 1.816763352670534e-06, "loss": 0.0051, "step": 20468 }, { "epoch": 6.35, "learning_rate": 1.816363272654531e-06, "loss": 0.0274, "step": 20469 }, { "epoch": 6.35, "learning_rate": 1.815963192638528e-06, "loss": 0.0022, "step": 20470 }, { "epoch": 6.35, "learning_rate": 1.8155631126225246e-06, "loss": 0.002, "step": 20471 }, { "epoch": 6.35, "learning_rate": 1.8151630326065215e-06, "loss": 0.0039, "step": 20472 }, { "epoch": 6.35, "learning_rate": 1.8147629525905184e-06, "loss": 0.0022, "step": 20473 }, { "epoch": 6.35, "learning_rate": 1.8143628725745149e-06, "loss": 0.0017, "step": 20474 }, { "epoch": 6.35, "learning_rate": 1.8139627925585118e-06, "loss": 0.0008, "step": 20475 }, { "epoch": 6.36, "learning_rate": 1.8135627125425087e-06, "loss": 0.0019, "step": 20476 }, { "epoch": 6.36, "learning_rate": 1.8131626325265053e-06, "loss": 0.0024, "step": 20477 }, { "epoch": 6.36, "learning_rate": 1.8127625525105022e-06, "loss": 0.0023, "step": 20478 }, { "epoch": 6.36, "learning_rate": 1.8123624724944991e-06, "loss": 0.0012, "step": 20479 }, { "epoch": 6.36, "learning_rate": 1.811962392478496e-06, "loss": 0.0028, "step": 20480 }, { "epoch": 6.36, "learning_rate": 1.8115623124624925e-06, "loss": 0.0017, "step": 20481 }, { "epoch": 6.36, "learning_rate": 1.8111622324464894e-06, "loss": 0.0011, "step": 20482 }, { "epoch": 6.36, "learning_rate": 1.8107621524304863e-06, "loss": 0.0009, "step": 20483 }, { "epoch": 6.36, "learning_rate": 1.810362072414483e-06, "loss": 0.0012, "step": 20484 }, { "epoch": 6.36, "learning_rate": 1.8099619923984799e-06, "loss": 0.0019, "step": 20485 }, { "epoch": 6.36, "learning_rate": 1.8095619123824768e-06, "loss": 0.0014, "step": 20486 }, { "epoch": 6.36, "learning_rate": 1.8091618323664733e-06, "loss": 0.0054, "step": 20487 }, { "epoch": 6.36, "learning_rate": 1.8087617523504702e-06, "loss": 0.001, "step": 20488 }, { "epoch": 6.36, "learning_rate": 1.808361672334467e-06, "loss": 0.0027, "step": 20489 }, { "epoch": 6.36, "learning_rate": 1.8079615923184637e-06, "loss": 0.002, "step": 20490 }, { "epoch": 6.36, "learning_rate": 1.8075615123024606e-06, "loss": 0.0013, "step": 20491 }, { "epoch": 6.36, "learning_rate": 1.8071614322864575e-06, "loss": 0.0218, "step": 20492 }, { "epoch": 6.36, "learning_rate": 1.8067613522704542e-06, "loss": 0.0044, "step": 20493 }, { "epoch": 6.36, "learning_rate": 1.806361272254451e-06, "loss": 0.0014, "step": 20494 }, { "epoch": 6.36, "learning_rate": 1.8059611922384478e-06, "loss": 0.0018, "step": 20495 }, { "epoch": 6.36, "learning_rate": 1.8055611122224445e-06, "loss": 0.0011, "step": 20496 }, { "epoch": 6.36, "learning_rate": 1.8051610322064414e-06, "loss": 0.0013, "step": 20497 }, { "epoch": 6.36, "learning_rate": 1.8047609521904383e-06, "loss": 0.0023, "step": 20498 }, { "epoch": 6.36, "learning_rate": 1.804360872174435e-06, "loss": 0.0089, "step": 20499 }, { "epoch": 6.36, "learning_rate": 1.8039607921584319e-06, "loss": 0.0017, "step": 20500 }, { "epoch": 6.36, "learning_rate": 1.8035607121424288e-06, "loss": 0.0055, "step": 20501 }, { "epoch": 6.36, "learning_rate": 1.8031606321264255e-06, "loss": 0.0062, "step": 20502 }, { "epoch": 6.36, "learning_rate": 1.8027605521104221e-06, "loss": 0.0021, "step": 20503 }, { "epoch": 6.36, "learning_rate": 1.802360472094419e-06, "loss": 0.0024, "step": 20504 }, { "epoch": 6.36, "learning_rate": 1.801960392078416e-06, "loss": 0.001, "step": 20505 }, { "epoch": 6.36, "learning_rate": 1.8015603120624126e-06, "loss": 0.0044, "step": 20506 }, { "epoch": 6.36, "learning_rate": 1.8011602320464095e-06, "loss": 0.0027, "step": 20507 }, { "epoch": 6.37, "learning_rate": 1.8007601520304064e-06, "loss": 0.0035, "step": 20508 }, { "epoch": 6.37, "learning_rate": 1.800360072014403e-06, "loss": 0.0036, "step": 20509 }, { "epoch": 6.37, "learning_rate": 1.7999599919983998e-06, "loss": 0.0015, "step": 20510 }, { "epoch": 6.37, "learning_rate": 1.7995599119823967e-06, "loss": 0.0021, "step": 20511 }, { "epoch": 6.37, "learning_rate": 1.7991598319663934e-06, "loss": 0.0027, "step": 20512 }, { "epoch": 6.37, "learning_rate": 1.7987597519503903e-06, "loss": 0.0018, "step": 20513 }, { "epoch": 6.37, "learning_rate": 1.7983596719343872e-06, "loss": 0.0015, "step": 20514 }, { "epoch": 6.37, "learning_rate": 1.7979595919183837e-06, "loss": 0.0013, "step": 20515 }, { "epoch": 6.37, "learning_rate": 1.7975595119023806e-06, "loss": 0.0016, "step": 20516 }, { "epoch": 6.37, "learning_rate": 1.7971594318863774e-06, "loss": 0.0023, "step": 20517 }, { "epoch": 6.37, "learning_rate": 1.7967593518703741e-06, "loss": 0.0028, "step": 20518 }, { "epoch": 6.37, "learning_rate": 1.796359271854371e-06, "loss": 0.0027, "step": 20519 }, { "epoch": 6.37, "learning_rate": 1.795959191838368e-06, "loss": 0.0011, "step": 20520 }, { "epoch": 6.37, "learning_rate": 1.7955591118223648e-06, "loss": 0.0033, "step": 20521 }, { "epoch": 6.37, "learning_rate": 1.7951590318063613e-06, "loss": 0.0036, "step": 20522 }, { "epoch": 6.37, "learning_rate": 1.7947589517903582e-06, "loss": 0.001, "step": 20523 }, { "epoch": 6.37, "learning_rate": 1.794358871774355e-06, "loss": 0.0014, "step": 20524 }, { "epoch": 6.37, "learning_rate": 1.7939587917583518e-06, "loss": 0.0037, "step": 20525 }, { "epoch": 6.37, "learning_rate": 1.7935587117423487e-06, "loss": 0.0403, "step": 20526 }, { "epoch": 6.37, "learning_rate": 1.7931586317263456e-06, "loss": 0.0024, "step": 20527 }, { "epoch": 6.37, "learning_rate": 1.792758551710342e-06, "loss": 0.0145, "step": 20528 }, { "epoch": 6.37, "learning_rate": 1.792358471694339e-06, "loss": 0.0009, "step": 20529 }, { "epoch": 6.37, "learning_rate": 1.7919583916783359e-06, "loss": 0.0014, "step": 20530 }, { "epoch": 6.37, "learning_rate": 1.7915583116623325e-06, "loss": 0.0013, "step": 20531 }, { "epoch": 6.37, "learning_rate": 1.7911582316463294e-06, "loss": 0.0056, "step": 20532 }, { "epoch": 6.37, "learning_rate": 1.7907581516303263e-06, "loss": 0.0015, "step": 20533 }, { "epoch": 6.37, "learning_rate": 1.7903580716143228e-06, "loss": 0.0012, "step": 20534 }, { "epoch": 6.37, "learning_rate": 1.7899579915983197e-06, "loss": 0.0348, "step": 20535 }, { "epoch": 6.37, "learning_rate": 1.7895579115823166e-06, "loss": 0.0037, "step": 20536 }, { "epoch": 6.37, "learning_rate": 1.7891578315663133e-06, "loss": 0.0016, "step": 20537 }, { "epoch": 6.37, "learning_rate": 1.7887577515503102e-06, "loss": 0.0011, "step": 20538 }, { "epoch": 6.37, "learning_rate": 1.788357671534307e-06, "loss": 0.0054, "step": 20539 }, { "epoch": 6.38, "learning_rate": 1.7879575915183038e-06, "loss": 0.0153, "step": 20540 }, { "epoch": 6.38, "learning_rate": 1.7875575115023005e-06, "loss": 0.0025, "step": 20541 }, { "epoch": 6.38, "learning_rate": 1.7871574314862974e-06, "loss": 0.0023, "step": 20542 }, { "epoch": 6.38, "learning_rate": 1.7867573514702943e-06, "loss": 0.0008, "step": 20543 }, { "epoch": 6.38, "learning_rate": 1.786357271454291e-06, "loss": 0.0069, "step": 20544 }, { "epoch": 6.38, "learning_rate": 1.7859571914382878e-06, "loss": 0.0048, "step": 20545 }, { "epoch": 6.38, "learning_rate": 1.7855571114222847e-06, "loss": 0.0342, "step": 20546 }, { "epoch": 6.38, "learning_rate": 1.7851570314062814e-06, "loss": 0.0016, "step": 20547 }, { "epoch": 6.38, "learning_rate": 1.7847569513902781e-06, "loss": 0.0152, "step": 20548 }, { "epoch": 6.38, "learning_rate": 1.784356871374275e-06, "loss": 0.0021, "step": 20549 }, { "epoch": 6.38, "learning_rate": 1.7839567913582717e-06, "loss": 0.0024, "step": 20550 }, { "epoch": 6.38, "learning_rate": 1.7835567113422686e-06, "loss": 0.0018, "step": 20551 }, { "epoch": 6.38, "learning_rate": 1.7831566313262655e-06, "loss": 0.0011, "step": 20552 }, { "epoch": 6.38, "learning_rate": 1.7827565513102622e-06, "loss": 0.0005, "step": 20553 }, { "epoch": 6.38, "learning_rate": 1.782356471294259e-06, "loss": 0.0027, "step": 20554 }, { "epoch": 6.38, "learning_rate": 1.781956391278256e-06, "loss": 0.0013, "step": 20555 }, { "epoch": 6.38, "learning_rate": 1.7815563112622524e-06, "loss": 0.002, "step": 20556 }, { "epoch": 6.38, "learning_rate": 1.7811562312462493e-06, "loss": 0.0011, "step": 20557 }, { "epoch": 6.38, "learning_rate": 1.7807561512302462e-06, "loss": 0.0265, "step": 20558 }, { "epoch": 6.38, "learning_rate": 1.780356071214243e-06, "loss": 0.0021, "step": 20559 }, { "epoch": 6.38, "learning_rate": 1.7799559911982398e-06, "loss": 0.0013, "step": 20560 }, { "epoch": 6.38, "learning_rate": 1.7795559111822367e-06, "loss": 0.0007, "step": 20561 }, { "epoch": 6.38, "learning_rate": 1.7791558311662336e-06, "loss": 0.0078, "step": 20562 }, { "epoch": 6.38, "learning_rate": 1.77875575115023e-06, "loss": 0.0014, "step": 20563 }, { "epoch": 6.38, "learning_rate": 1.778355671134227e-06, "loss": 0.0023, "step": 20564 }, { "epoch": 6.38, "learning_rate": 1.777955591118224e-06, "loss": 0.0021, "step": 20565 }, { "epoch": 6.38, "learning_rate": 1.7775555111022206e-06, "loss": 0.001, "step": 20566 }, { "epoch": 6.38, "learning_rate": 1.7771554310862175e-06, "loss": 0.002, "step": 20567 }, { "epoch": 6.38, "learning_rate": 1.7767553510702144e-06, "loss": 0.0011, "step": 20568 }, { "epoch": 6.38, "learning_rate": 1.7763552710542109e-06, "loss": 0.0019, "step": 20569 }, { "epoch": 6.38, "learning_rate": 1.7759551910382078e-06, "loss": 0.0024, "step": 20570 }, { "epoch": 6.38, "learning_rate": 1.7755551110222046e-06, "loss": 0.002, "step": 20571 }, { "epoch": 6.39, "learning_rate": 1.7751550310062013e-06, "loss": 0.0014, "step": 20572 }, { "epoch": 6.39, "learning_rate": 1.7747549509901982e-06, "loss": 0.0013, "step": 20573 }, { "epoch": 6.39, "learning_rate": 1.7743548709741951e-06, "loss": 0.0035, "step": 20574 }, { "epoch": 6.39, "learning_rate": 1.7739547909581916e-06, "loss": 0.0009, "step": 20575 }, { "epoch": 6.39, "learning_rate": 1.7735547109421885e-06, "loss": 0.0016, "step": 20576 }, { "epoch": 6.39, "learning_rate": 1.7731546309261854e-06, "loss": 0.0016, "step": 20577 }, { "epoch": 6.39, "learning_rate": 1.772754550910182e-06, "loss": 0.0012, "step": 20578 }, { "epoch": 6.39, "learning_rate": 1.772354470894179e-06, "loss": 0.0369, "step": 20579 }, { "epoch": 6.39, "learning_rate": 1.7719543908781759e-06, "loss": 0.0008, "step": 20580 }, { "epoch": 6.39, "learning_rate": 1.7715543108621724e-06, "loss": 0.0035, "step": 20581 }, { "epoch": 6.39, "learning_rate": 1.7711542308461693e-06, "loss": 0.001, "step": 20582 }, { "epoch": 6.39, "learning_rate": 1.7707541508301662e-06, "loss": 0.0033, "step": 20583 }, { "epoch": 6.39, "learning_rate": 1.770354070814163e-06, "loss": 0.0014, "step": 20584 }, { "epoch": 6.39, "learning_rate": 1.7699539907981597e-06, "loss": 0.0017, "step": 20585 }, { "epoch": 6.39, "learning_rate": 1.7695539107821566e-06, "loss": 0.0032, "step": 20586 }, { "epoch": 6.39, "learning_rate": 1.7691538307661535e-06, "loss": 0.002, "step": 20587 }, { "epoch": 6.39, "learning_rate": 1.76875375075015e-06, "loss": 0.0071, "step": 20588 }, { "epoch": 6.39, "learning_rate": 1.768353670734147e-06, "loss": 0.0012, "step": 20589 }, { "epoch": 6.39, "learning_rate": 1.7679535907181438e-06, "loss": 0.0027, "step": 20590 }, { "epoch": 6.39, "learning_rate": 1.7675535107021405e-06, "loss": 0.0031, "step": 20591 }, { "epoch": 6.39, "learning_rate": 1.7671534306861374e-06, "loss": 0.0011, "step": 20592 }, { "epoch": 6.39, "learning_rate": 1.7667533506701343e-06, "loss": 0.0017, "step": 20593 }, { "epoch": 6.39, "learning_rate": 1.7663532706541308e-06, "loss": 0.0016, "step": 20594 }, { "epoch": 6.39, "learning_rate": 1.7659531906381277e-06, "loss": 0.002, "step": 20595 }, { "epoch": 6.39, "learning_rate": 1.7655531106221246e-06, "loss": 0.0185, "step": 20596 }, { "epoch": 6.39, "learning_rate": 1.7651530306061212e-06, "loss": 0.0013, "step": 20597 }, { "epoch": 6.39, "learning_rate": 1.7647529505901181e-06, "loss": 0.0033, "step": 20598 }, { "epoch": 6.39, "learning_rate": 1.764352870574115e-06, "loss": 0.0068, "step": 20599 }, { "epoch": 6.39, "learning_rate": 1.7639527905581117e-06, "loss": 0.0027, "step": 20600 }, { "epoch": 6.39, "learning_rate": 1.7635527105421086e-06, "loss": 0.0024, "step": 20601 }, { "epoch": 6.39, "learning_rate": 1.7631526305261053e-06, "loss": 0.0033, "step": 20602 }, { "epoch": 6.39, "learning_rate": 1.7627525505101022e-06, "loss": 0.0037, "step": 20603 }, { "epoch": 6.4, "learning_rate": 1.762352470494099e-06, "loss": 0.002, "step": 20604 }, { "epoch": 6.4, "learning_rate": 1.7619523904780958e-06, "loss": 0.0021, "step": 20605 }, { "epoch": 6.4, "learning_rate": 1.7615523104620927e-06, "loss": 0.0018, "step": 20606 }, { "epoch": 6.4, "learning_rate": 1.7611522304460894e-06, "loss": 0.0035, "step": 20607 }, { "epoch": 6.4, "learning_rate": 1.7607521504300863e-06, "loss": 0.0145, "step": 20608 }, { "epoch": 6.4, "learning_rate": 1.7603520704140832e-06, "loss": 0.0017, "step": 20609 }, { "epoch": 6.4, "learning_rate": 1.7599519903980796e-06, "loss": 0.006, "step": 20610 }, { "epoch": 6.4, "learning_rate": 1.7595519103820765e-06, "loss": 0.0056, "step": 20611 }, { "epoch": 6.4, "learning_rate": 1.7591518303660734e-06, "loss": 0.002, "step": 20612 }, { "epoch": 6.4, "learning_rate": 1.7587517503500701e-06, "loss": 0.0011, "step": 20613 }, { "epoch": 6.4, "learning_rate": 1.758351670334067e-06, "loss": 0.0111, "step": 20614 }, { "epoch": 6.4, "learning_rate": 1.757951590318064e-06, "loss": 0.0013, "step": 20615 }, { "epoch": 6.4, "learning_rate": 1.7575515103020604e-06, "loss": 0.0009, "step": 20616 }, { "epoch": 6.4, "learning_rate": 1.7571514302860573e-06, "loss": 0.0009, "step": 20617 }, { "epoch": 6.4, "learning_rate": 1.7567513502700542e-06, "loss": 0.0009, "step": 20618 }, { "epoch": 6.4, "learning_rate": 1.7563512702540509e-06, "loss": 0.0009, "step": 20619 }, { "epoch": 6.4, "learning_rate": 1.7559511902380478e-06, "loss": 0.0018, "step": 20620 }, { "epoch": 6.4, "learning_rate": 1.7555511102220447e-06, "loss": 0.0039, "step": 20621 }, { "epoch": 6.4, "learning_rate": 1.7551510302060412e-06, "loss": 0.0282, "step": 20622 }, { "epoch": 6.4, "learning_rate": 1.754750950190038e-06, "loss": 0.0021, "step": 20623 }, { "epoch": 6.4, "learning_rate": 1.754350870174035e-06, "loss": 0.0011, "step": 20624 }, { "epoch": 6.4, "learning_rate": 1.7539507901580318e-06, "loss": 0.001, "step": 20625 }, { "epoch": 6.4, "learning_rate": 1.7535507101420285e-06, "loss": 0.0021, "step": 20626 }, { "epoch": 6.4, "learning_rate": 1.7531506301260254e-06, "loss": 0.0035, "step": 20627 }, { "epoch": 6.4, "learning_rate": 1.7527505501100223e-06, "loss": 0.0016, "step": 20628 }, { "epoch": 6.4, "learning_rate": 1.7523504700940188e-06, "loss": 0.001, "step": 20629 }, { "epoch": 6.4, "learning_rate": 1.7519503900780157e-06, "loss": 0.0025, "step": 20630 }, { "epoch": 6.4, "learning_rate": 1.7515503100620126e-06, "loss": 0.0011, "step": 20631 }, { "epoch": 6.4, "learning_rate": 1.7511502300460093e-06, "loss": 0.0046, "step": 20632 }, { "epoch": 6.4, "learning_rate": 1.7507501500300062e-06, "loss": 0.0012, "step": 20633 }, { "epoch": 6.4, "learning_rate": 1.750350070014003e-06, "loss": 0.0025, "step": 20634 }, { "epoch": 6.4, "learning_rate": 1.7499499899979996e-06, "loss": 0.0046, "step": 20635 }, { "epoch": 6.4, "learning_rate": 1.7495499099819965e-06, "loss": 0.0006, "step": 20636 }, { "epoch": 6.41, "learning_rate": 1.7491498299659934e-06, "loss": 0.001, "step": 20637 }, { "epoch": 6.41, "learning_rate": 1.74874974994999e-06, "loss": 0.003, "step": 20638 }, { "epoch": 6.41, "learning_rate": 1.748349669933987e-06, "loss": 0.0118, "step": 20639 }, { "epoch": 6.41, "learning_rate": 1.7479495899179838e-06, "loss": 0.0026, "step": 20640 }, { "epoch": 6.41, "learning_rate": 1.7475495099019803e-06, "loss": 0.0057, "step": 20641 }, { "epoch": 6.41, "learning_rate": 1.7471494298859772e-06, "loss": 0.0226, "step": 20642 }, { "epoch": 6.41, "learning_rate": 1.7467493498699741e-06, "loss": 0.0024, "step": 20643 }, { "epoch": 6.41, "learning_rate": 1.746349269853971e-06, "loss": 0.0022, "step": 20644 }, { "epoch": 6.41, "learning_rate": 1.7459491898379677e-06, "loss": 0.0011, "step": 20645 }, { "epoch": 6.41, "learning_rate": 1.7455491098219646e-06, "loss": 0.0052, "step": 20646 }, { "epoch": 6.41, "learning_rate": 1.7451490298059615e-06, "loss": 0.0084, "step": 20647 }, { "epoch": 6.41, "learning_rate": 1.744748949789958e-06, "loss": 0.0109, "step": 20648 }, { "epoch": 6.41, "learning_rate": 1.7443488697739549e-06, "loss": 0.0012, "step": 20649 }, { "epoch": 6.41, "learning_rate": 1.7439487897579518e-06, "loss": 0.0031, "step": 20650 }, { "epoch": 6.41, "learning_rate": 1.7435487097419484e-06, "loss": 0.0068, "step": 20651 }, { "epoch": 6.41, "learning_rate": 1.7431486297259453e-06, "loss": 0.0022, "step": 20652 }, { "epoch": 6.41, "learning_rate": 1.7427485497099422e-06, "loss": 0.0062, "step": 20653 }, { "epoch": 6.41, "learning_rate": 1.742348469693939e-06, "loss": 0.0031, "step": 20654 }, { "epoch": 6.41, "learning_rate": 1.7419483896779358e-06, "loss": 0.001, "step": 20655 }, { "epoch": 6.41, "learning_rate": 1.7415483096619325e-06, "loss": 0.0013, "step": 20656 }, { "epoch": 6.41, "learning_rate": 1.7411482296459292e-06, "loss": 0.0096, "step": 20657 }, { "epoch": 6.41, "learning_rate": 1.740748149629926e-06, "loss": 0.0013, "step": 20658 }, { "epoch": 6.41, "learning_rate": 1.740348069613923e-06, "loss": 0.0011, "step": 20659 }, { "epoch": 6.41, "learning_rate": 1.7399479895979197e-06, "loss": 0.0942, "step": 20660 }, { "epoch": 6.41, "learning_rate": 1.7395479095819166e-06, "loss": 0.0027, "step": 20661 }, { "epoch": 6.41, "learning_rate": 1.7391478295659135e-06, "loss": 0.0034, "step": 20662 }, { "epoch": 6.41, "learning_rate": 1.73874774954991e-06, "loss": 0.0051, "step": 20663 }, { "epoch": 6.41, "learning_rate": 1.7383476695339068e-06, "loss": 0.001, "step": 20664 }, { "epoch": 6.41, "learning_rate": 1.7379475895179037e-06, "loss": 0.0504, "step": 20665 }, { "epoch": 6.41, "learning_rate": 1.7375475095019006e-06, "loss": 0.0017, "step": 20666 }, { "epoch": 6.41, "learning_rate": 1.7371474294858973e-06, "loss": 0.0095, "step": 20667 }, { "epoch": 6.41, "learning_rate": 1.7367473494698942e-06, "loss": 0.0016, "step": 20668 }, { "epoch": 6.42, "learning_rate": 1.7363472694538911e-06, "loss": 0.0019, "step": 20669 }, { "epoch": 6.42, "learning_rate": 1.7359471894378876e-06, "loss": 0.0016, "step": 20670 }, { "epoch": 6.42, "learning_rate": 1.7355471094218845e-06, "loss": 0.0013, "step": 20671 }, { "epoch": 6.42, "learning_rate": 1.7351470294058814e-06, "loss": 0.0014, "step": 20672 }, { "epoch": 6.42, "learning_rate": 1.734746949389878e-06, "loss": 0.0014, "step": 20673 }, { "epoch": 6.42, "learning_rate": 1.734346869373875e-06, "loss": 0.0009, "step": 20674 }, { "epoch": 6.42, "learning_rate": 1.7339467893578719e-06, "loss": 0.0095, "step": 20675 }, { "epoch": 6.42, "learning_rate": 1.7335467093418684e-06, "loss": 0.003, "step": 20676 }, { "epoch": 6.42, "learning_rate": 1.7331466293258653e-06, "loss": 0.0023, "step": 20677 }, { "epoch": 6.42, "learning_rate": 1.7327465493098622e-06, "loss": 0.0014, "step": 20678 }, { "epoch": 6.42, "learning_rate": 1.7323464692938588e-06, "loss": 0.0295, "step": 20679 }, { "epoch": 6.42, "learning_rate": 1.7319463892778557e-06, "loss": 0.0098, "step": 20680 }, { "epoch": 6.42, "learning_rate": 1.7315463092618526e-06, "loss": 0.002, "step": 20681 }, { "epoch": 6.42, "learning_rate": 1.7311462292458491e-06, "loss": 0.0039, "step": 20682 }, { "epoch": 6.42, "learning_rate": 1.730746149229846e-06, "loss": 0.0019, "step": 20683 }, { "epoch": 6.42, "learning_rate": 1.730346069213843e-06, "loss": 0.0071, "step": 20684 }, { "epoch": 6.42, "learning_rate": 1.7299459891978398e-06, "loss": 0.0017, "step": 20685 }, { "epoch": 6.42, "learning_rate": 1.7295459091818365e-06, "loss": 0.0016, "step": 20686 }, { "epoch": 6.42, "learning_rate": 1.7291458291658334e-06, "loss": 0.0017, "step": 20687 }, { "epoch": 6.42, "learning_rate": 1.7287457491498303e-06, "loss": 0.0011, "step": 20688 }, { "epoch": 6.42, "learning_rate": 1.7283456691338268e-06, "loss": 0.0034, "step": 20689 }, { "epoch": 6.42, "learning_rate": 1.7279455891178237e-06, "loss": 0.0042, "step": 20690 }, { "epoch": 6.42, "learning_rate": 1.7275455091018206e-06, "loss": 0.0172, "step": 20691 }, { "epoch": 6.42, "learning_rate": 1.7271454290858172e-06, "loss": 0.0155, "step": 20692 }, { "epoch": 6.42, "learning_rate": 1.7267453490698141e-06, "loss": 0.0011, "step": 20693 }, { "epoch": 6.42, "learning_rate": 1.726345269053811e-06, "loss": 0.0031, "step": 20694 }, { "epoch": 6.42, "learning_rate": 1.7259451890378075e-06, "loss": 0.0025, "step": 20695 }, { "epoch": 6.42, "learning_rate": 1.7255451090218044e-06, "loss": 0.0021, "step": 20696 }, { "epoch": 6.42, "learning_rate": 1.7251450290058013e-06, "loss": 0.0013, "step": 20697 }, { "epoch": 6.42, "learning_rate": 1.724744948989798e-06, "loss": 0.0012, "step": 20698 }, { "epoch": 6.42, "learning_rate": 1.7243448689737949e-06, "loss": 0.0017, "step": 20699 }, { "epoch": 6.42, "learning_rate": 1.7239447889577918e-06, "loss": 0.005, "step": 20700 }, { "epoch": 6.43, "learning_rate": 1.7235447089417885e-06, "loss": 0.0007, "step": 20701 }, { "epoch": 6.43, "learning_rate": 1.7231446289257852e-06, "loss": 0.001, "step": 20702 }, { "epoch": 6.43, "learning_rate": 1.722744548909782e-06, "loss": 0.0021, "step": 20703 }, { "epoch": 6.43, "learning_rate": 1.7223444688937787e-06, "loss": 0.0034, "step": 20704 }, { "epoch": 6.43, "learning_rate": 1.7219443888777756e-06, "loss": 0.0012, "step": 20705 }, { "epoch": 6.43, "learning_rate": 1.7215443088617725e-06, "loss": 0.001, "step": 20706 }, { "epoch": 6.43, "learning_rate": 1.7211442288457694e-06, "loss": 0.0018, "step": 20707 }, { "epoch": 6.43, "learning_rate": 1.7207441488297661e-06, "loss": 0.0401, "step": 20708 }, { "epoch": 6.43, "learning_rate": 1.720344068813763e-06, "loss": 0.0006, "step": 20709 }, { "epoch": 6.43, "learning_rate": 1.7199439887977597e-06, "loss": 0.0023, "step": 20710 }, { "epoch": 6.43, "learning_rate": 1.7195439087817564e-06, "loss": 0.0053, "step": 20711 }, { "epoch": 6.43, "learning_rate": 1.7191438287657533e-06, "loss": 0.0053, "step": 20712 }, { "epoch": 6.43, "learning_rate": 1.7187437487497502e-06, "loss": 0.0023, "step": 20713 }, { "epoch": 6.43, "learning_rate": 1.7183436687337469e-06, "loss": 0.0019, "step": 20714 }, { "epoch": 6.43, "learning_rate": 1.7179435887177438e-06, "loss": 0.0062, "step": 20715 }, { "epoch": 6.43, "learning_rate": 1.7175435087017407e-06, "loss": 0.0014, "step": 20716 }, { "epoch": 6.43, "learning_rate": 1.7171434286857372e-06, "loss": 0.0013, "step": 20717 }, { "epoch": 6.43, "learning_rate": 1.716743348669734e-06, "loss": 0.0014, "step": 20718 }, { "epoch": 6.43, "learning_rate": 1.716343268653731e-06, "loss": 0.0007, "step": 20719 }, { "epoch": 6.43, "learning_rate": 1.7159431886377276e-06, "loss": 0.0032, "step": 20720 }, { "epoch": 6.43, "learning_rate": 1.7155431086217245e-06, "loss": 0.0022, "step": 20721 }, { "epoch": 6.43, "learning_rate": 1.7151430286057214e-06, "loss": 0.0055, "step": 20722 }, { "epoch": 6.43, "learning_rate": 1.714742948589718e-06, "loss": 0.0017, "step": 20723 }, { "epoch": 6.43, "learning_rate": 1.7143428685737148e-06, "loss": 0.0012, "step": 20724 }, { "epoch": 6.43, "learning_rate": 1.7139427885577117e-06, "loss": 0.0119, "step": 20725 }, { "epoch": 6.43, "learning_rate": 1.7135427085417084e-06, "loss": 0.005, "step": 20726 }, { "epoch": 6.43, "learning_rate": 1.7131426285257053e-06, "loss": 0.0016, "step": 20727 }, { "epoch": 6.43, "learning_rate": 1.7127425485097022e-06, "loss": 0.0026, "step": 20728 }, { "epoch": 6.43, "learning_rate": 1.712342468493699e-06, "loss": 0.0014, "step": 20729 }, { "epoch": 6.43, "learning_rate": 1.7119423884776956e-06, "loss": 0.0022, "step": 20730 }, { "epoch": 6.43, "learning_rate": 1.7115423084616925e-06, "loss": 0.0013, "step": 20731 }, { "epoch": 6.43, "learning_rate": 1.7111422284456894e-06, "loss": 0.0019, "step": 20732 }, { "epoch": 6.44, "learning_rate": 1.710742148429686e-06, "loss": 0.0018, "step": 20733 }, { "epoch": 6.44, "learning_rate": 1.710342068413683e-06, "loss": 0.0028, "step": 20734 }, { "epoch": 6.44, "learning_rate": 1.7099419883976798e-06, "loss": 0.0008, "step": 20735 }, { "epoch": 6.44, "learning_rate": 1.7095419083816763e-06, "loss": 0.0097, "step": 20736 }, { "epoch": 6.44, "learning_rate": 1.7091418283656732e-06, "loss": 0.0045, "step": 20737 }, { "epoch": 6.44, "learning_rate": 1.70874174834967e-06, "loss": 0.0052, "step": 20738 }, { "epoch": 6.44, "learning_rate": 1.7083416683336668e-06, "loss": 0.0012, "step": 20739 }, { "epoch": 6.44, "learning_rate": 1.7079415883176637e-06, "loss": 0.0007, "step": 20740 }, { "epoch": 6.44, "learning_rate": 1.7075415083016606e-06, "loss": 0.0049, "step": 20741 }, { "epoch": 6.44, "learning_rate": 1.707141428285657e-06, "loss": 0.0014, "step": 20742 }, { "epoch": 6.44, "learning_rate": 1.706741348269654e-06, "loss": 0.0012, "step": 20743 }, { "epoch": 6.44, "learning_rate": 1.7063412682536509e-06, "loss": 0.0057, "step": 20744 }, { "epoch": 6.44, "learning_rate": 1.7059411882376475e-06, "loss": 0.0026, "step": 20745 }, { "epoch": 6.44, "learning_rate": 1.7055411082216444e-06, "loss": 0.0031, "step": 20746 }, { "epoch": 6.44, "learning_rate": 1.7051410282056413e-06, "loss": 0.0012, "step": 20747 }, { "epoch": 6.44, "learning_rate": 1.7047409481896382e-06, "loss": 0.0012, "step": 20748 }, { "epoch": 6.44, "learning_rate": 1.7043408681736347e-06, "loss": 0.0024, "step": 20749 }, { "epoch": 6.44, "learning_rate": 1.7039407881576316e-06, "loss": 0.0013, "step": 20750 }, { "epoch": 6.44, "learning_rate": 1.7035407081416285e-06, "loss": 0.0035, "step": 20751 }, { "epoch": 6.44, "learning_rate": 1.7031406281256252e-06, "loss": 0.0019, "step": 20752 }, { "epoch": 6.44, "learning_rate": 1.702740548109622e-06, "loss": 0.0019, "step": 20753 }, { "epoch": 6.44, "learning_rate": 1.702340468093619e-06, "loss": 0.0372, "step": 20754 }, { "epoch": 6.44, "learning_rate": 1.7019403880776157e-06, "loss": 0.0008, "step": 20755 }, { "epoch": 6.44, "learning_rate": 1.7015403080616124e-06, "loss": 0.0032, "step": 20756 }, { "epoch": 6.44, "learning_rate": 1.7011402280456093e-06, "loss": 0.0019, "step": 20757 }, { "epoch": 6.44, "learning_rate": 1.700740148029606e-06, "loss": 0.0037, "step": 20758 }, { "epoch": 6.44, "learning_rate": 1.7003400680136028e-06, "loss": 0.0009, "step": 20759 }, { "epoch": 6.44, "learning_rate": 1.6999399879975997e-06, "loss": 0.0289, "step": 20760 }, { "epoch": 6.44, "learning_rate": 1.6995399079815964e-06, "loss": 0.0019, "step": 20761 }, { "epoch": 6.44, "learning_rate": 1.6991398279655933e-06, "loss": 0.0035, "step": 20762 }, { "epoch": 6.44, "learning_rate": 1.6987397479495902e-06, "loss": 0.0164, "step": 20763 }, { "epoch": 6.44, "learning_rate": 1.6983396679335867e-06, "loss": 0.0083, "step": 20764 }, { "epoch": 6.45, "learning_rate": 1.6979395879175836e-06, "loss": 0.0189, "step": 20765 }, { "epoch": 6.45, "learning_rate": 1.6975395079015805e-06, "loss": 0.0026, "step": 20766 }, { "epoch": 6.45, "learning_rate": 1.6971394278855772e-06, "loss": 0.01, "step": 20767 }, { "epoch": 6.45, "learning_rate": 1.696739347869574e-06, "loss": 0.0008, "step": 20768 }, { "epoch": 6.45, "learning_rate": 1.696339267853571e-06, "loss": 0.0008, "step": 20769 }, { "epoch": 6.45, "learning_rate": 1.6959391878375679e-06, "loss": 0.0006, "step": 20770 }, { "epoch": 6.45, "learning_rate": 1.6955391078215644e-06, "loss": 0.0039, "step": 20771 }, { "epoch": 6.45, "learning_rate": 1.6951390278055612e-06, "loss": 0.0014, "step": 20772 }, { "epoch": 6.45, "learning_rate": 1.6947389477895581e-06, "loss": 0.0015, "step": 20773 }, { "epoch": 6.45, "learning_rate": 1.6943388677735548e-06, "loss": 0.008, "step": 20774 }, { "epoch": 6.45, "learning_rate": 1.6939387877575517e-06, "loss": 0.0012, "step": 20775 }, { "epoch": 6.45, "learning_rate": 1.6935387077415486e-06, "loss": 0.0017, "step": 20776 }, { "epoch": 6.45, "learning_rate": 1.693138627725545e-06, "loss": 0.0009, "step": 20777 }, { "epoch": 6.45, "learning_rate": 1.692738547709542e-06, "loss": 0.0011, "step": 20778 }, { "epoch": 6.45, "learning_rate": 1.692338467693539e-06, "loss": 0.019, "step": 20779 }, { "epoch": 6.45, "learning_rate": 1.6919383876775356e-06, "loss": 0.0008, "step": 20780 }, { "epoch": 6.45, "learning_rate": 1.6915383076615325e-06, "loss": 0.0083, "step": 20781 }, { "epoch": 6.45, "learning_rate": 1.6911382276455294e-06, "loss": 0.007, "step": 20782 }, { "epoch": 6.45, "learning_rate": 1.6907381476295259e-06, "loss": 0.0035, "step": 20783 }, { "epoch": 6.45, "learning_rate": 1.6903380676135228e-06, "loss": 0.0022, "step": 20784 }, { "epoch": 6.45, "learning_rate": 1.6899379875975197e-06, "loss": 0.0041, "step": 20785 }, { "epoch": 6.45, "learning_rate": 1.6895379075815163e-06, "loss": 0.0015, "step": 20786 }, { "epoch": 6.45, "learning_rate": 1.6891378275655132e-06, "loss": 0.0245, "step": 20787 }, { "epoch": 6.45, "learning_rate": 1.6887377475495101e-06, "loss": 0.0017, "step": 20788 }, { "epoch": 6.45, "learning_rate": 1.688337667533507e-06, "loss": 0.0093, "step": 20789 }, { "epoch": 6.45, "learning_rate": 1.6879375875175035e-06, "loss": 0.0058, "step": 20790 }, { "epoch": 6.45, "learning_rate": 1.6875375075015004e-06, "loss": 0.0015, "step": 20791 }, { "epoch": 6.45, "learning_rate": 1.6871374274854973e-06, "loss": 0.0023, "step": 20792 }, { "epoch": 6.45, "learning_rate": 1.686737347469494e-06, "loss": 0.0039, "step": 20793 }, { "epoch": 6.45, "learning_rate": 1.6863372674534909e-06, "loss": 0.0011, "step": 20794 }, { "epoch": 6.45, "learning_rate": 1.6859371874374878e-06, "loss": 0.0013, "step": 20795 }, { "epoch": 6.45, "learning_rate": 1.6855371074214843e-06, "loss": 0.0007, "step": 20796 }, { "epoch": 6.45, "learning_rate": 1.6851370274054812e-06, "loss": 0.0022, "step": 20797 }, { "epoch": 6.46, "learning_rate": 1.684736947389478e-06, "loss": 0.0017, "step": 20798 }, { "epoch": 6.46, "learning_rate": 1.6843368673734747e-06, "loss": 0.0013, "step": 20799 }, { "epoch": 6.46, "learning_rate": 1.6839367873574716e-06, "loss": 0.0014, "step": 20800 }, { "epoch": 6.46, "learning_rate": 1.6835367073414685e-06, "loss": 0.0033, "step": 20801 }, { "epoch": 6.46, "learning_rate": 1.6831366273254652e-06, "loss": 0.0012, "step": 20802 }, { "epoch": 6.46, "learning_rate": 1.682736547309462e-06, "loss": 0.0007, "step": 20803 }, { "epoch": 6.46, "learning_rate": 1.6823364672934588e-06, "loss": 0.0013, "step": 20804 }, { "epoch": 6.46, "learning_rate": 1.6819363872774555e-06, "loss": 0.004, "step": 20805 }, { "epoch": 6.46, "learning_rate": 1.6815363072614524e-06, "loss": 0.0012, "step": 20806 }, { "epoch": 6.46, "learning_rate": 1.6811362272454493e-06, "loss": 0.0018, "step": 20807 }, { "epoch": 6.46, "learning_rate": 1.680736147229446e-06, "loss": 0.0026, "step": 20808 }, { "epoch": 6.46, "learning_rate": 1.6803360672134429e-06, "loss": 0.0017, "step": 20809 }, { "epoch": 6.46, "learning_rate": 1.6799359871974396e-06, "loss": 0.0023, "step": 20810 }, { "epoch": 6.46, "learning_rate": 1.6795359071814365e-06, "loss": 0.0011, "step": 20811 }, { "epoch": 6.46, "learning_rate": 1.6791358271654331e-06, "loss": 0.0012, "step": 20812 }, { "epoch": 6.46, "learning_rate": 1.67873574714943e-06, "loss": 0.002, "step": 20813 }, { "epoch": 6.46, "learning_rate": 1.678335667133427e-06, "loss": 0.002, "step": 20814 }, { "epoch": 6.46, "learning_rate": 1.6779355871174236e-06, "loss": 0.0055, "step": 20815 }, { "epoch": 6.46, "learning_rate": 1.6775355071014205e-06, "loss": 0.0029, "step": 20816 }, { "epoch": 6.46, "learning_rate": 1.6771354270854174e-06, "loss": 0.0085, "step": 20817 }, { "epoch": 6.46, "learning_rate": 1.676735347069414e-06, "loss": 0.001, "step": 20818 }, { "epoch": 6.46, "learning_rate": 1.6763352670534108e-06, "loss": 0.0014, "step": 20819 }, { "epoch": 6.46, "learning_rate": 1.6759351870374077e-06, "loss": 0.0035, "step": 20820 }, { "epoch": 6.46, "learning_rate": 1.6755351070214044e-06, "loss": 0.0012, "step": 20821 }, { "epoch": 6.46, "learning_rate": 1.6751350270054013e-06, "loss": 0.0013, "step": 20822 }, { "epoch": 6.46, "learning_rate": 1.6747349469893982e-06, "loss": 0.0161, "step": 20823 }, { "epoch": 6.46, "learning_rate": 1.6743348669733947e-06, "loss": 0.0049, "step": 20824 }, { "epoch": 6.46, "learning_rate": 1.6739347869573916e-06, "loss": 0.0081, "step": 20825 }, { "epoch": 6.46, "learning_rate": 1.6735347069413884e-06, "loss": 0.0041, "step": 20826 }, { "epoch": 6.46, "learning_rate": 1.6731346269253851e-06, "loss": 0.0016, "step": 20827 }, { "epoch": 6.46, "learning_rate": 1.672734546909382e-06, "loss": 0.002, "step": 20828 }, { "epoch": 6.46, "learning_rate": 1.672334466893379e-06, "loss": 0.0089, "step": 20829 }, { "epoch": 6.47, "learning_rate": 1.6719343868773758e-06, "loss": 0.0011, "step": 20830 }, { "epoch": 6.47, "learning_rate": 1.6715343068613723e-06, "loss": 0.002, "step": 20831 }, { "epoch": 6.47, "learning_rate": 1.6711342268453692e-06, "loss": 0.0011, "step": 20832 }, { "epoch": 6.47, "learning_rate": 1.670734146829366e-06, "loss": 0.0116, "step": 20833 }, { "epoch": 6.47, "learning_rate": 1.6703340668133628e-06, "loss": 0.0011, "step": 20834 }, { "epoch": 6.47, "learning_rate": 1.6699339867973597e-06, "loss": 0.0023, "step": 20835 }, { "epoch": 6.47, "learning_rate": 1.6695339067813566e-06, "loss": 0.0047, "step": 20836 }, { "epoch": 6.47, "learning_rate": 1.669133826765353e-06, "loss": 0.0016, "step": 20837 }, { "epoch": 6.47, "learning_rate": 1.66873374674935e-06, "loss": 0.0225, "step": 20838 }, { "epoch": 6.47, "learning_rate": 1.6683336667333469e-06, "loss": 0.0014, "step": 20839 }, { "epoch": 6.47, "learning_rate": 1.6679335867173435e-06, "loss": 0.0028, "step": 20840 }, { "epoch": 6.47, "learning_rate": 1.6675335067013404e-06, "loss": 0.0012, "step": 20841 }, { "epoch": 6.47, "learning_rate": 1.6671334266853373e-06, "loss": 0.001, "step": 20842 }, { "epoch": 6.47, "learning_rate": 1.6667333466693338e-06, "loss": 0.0036, "step": 20843 }, { "epoch": 6.47, "learning_rate": 1.6663332666533307e-06, "loss": 0.0013, "step": 20844 }, { "epoch": 6.47, "learning_rate": 1.6659331866373276e-06, "loss": 0.0077, "step": 20845 }, { "epoch": 6.47, "learning_rate": 1.6655331066213243e-06, "loss": 0.0023, "step": 20846 }, { "epoch": 6.47, "learning_rate": 1.6651330266053212e-06, "loss": 0.0506, "step": 20847 }, { "epoch": 6.47, "learning_rate": 1.664732946589318e-06, "loss": 0.0013, "step": 20848 }, { "epoch": 6.47, "learning_rate": 1.6643328665733146e-06, "loss": 0.0026, "step": 20849 }, { "epoch": 6.47, "learning_rate": 1.6639327865573115e-06, "loss": 0.0038, "step": 20850 }, { "epoch": 6.47, "learning_rate": 1.6635327065413084e-06, "loss": 0.001, "step": 20851 }, { "epoch": 6.47, "learning_rate": 1.6631326265253053e-06, "loss": 0.0015, "step": 20852 }, { "epoch": 6.47, "learning_rate": 1.662732546509302e-06, "loss": 0.0015, "step": 20853 }, { "epoch": 6.47, "learning_rate": 1.6623324664932988e-06, "loss": 0.0041, "step": 20854 }, { "epoch": 6.47, "learning_rate": 1.6619323864772957e-06, "loss": 0.0457, "step": 20855 }, { "epoch": 6.47, "learning_rate": 1.6615323064612924e-06, "loss": 0.0017, "step": 20856 }, { "epoch": 6.47, "learning_rate": 1.6611322264452891e-06, "loss": 0.0031, "step": 20857 }, { "epoch": 6.47, "learning_rate": 1.660732146429286e-06, "loss": 0.0019, "step": 20858 }, { "epoch": 6.47, "learning_rate": 1.6603320664132827e-06, "loss": 0.0019, "step": 20859 }, { "epoch": 6.47, "learning_rate": 1.6599319863972796e-06, "loss": 0.0016, "step": 20860 }, { "epoch": 6.47, "learning_rate": 1.6595319063812765e-06, "loss": 0.0037, "step": 20861 }, { "epoch": 6.48, "learning_rate": 1.6591318263652732e-06, "loss": 0.0021, "step": 20862 }, { "epoch": 6.48, "learning_rate": 1.65873174634927e-06, "loss": 0.0017, "step": 20863 }, { "epoch": 6.48, "learning_rate": 1.6583316663332668e-06, "loss": 0.0104, "step": 20864 }, { "epoch": 6.48, "learning_rate": 1.6579315863172634e-06, "loss": 0.0012, "step": 20865 }, { "epoch": 6.48, "learning_rate": 1.6575315063012603e-06, "loss": 0.0022, "step": 20866 }, { "epoch": 6.48, "learning_rate": 1.6571314262852572e-06, "loss": 0.0012, "step": 20867 }, { "epoch": 6.48, "learning_rate": 1.656731346269254e-06, "loss": 0.0024, "step": 20868 }, { "epoch": 6.48, "learning_rate": 1.6563312662532508e-06, "loss": 0.0013, "step": 20869 }, { "epoch": 6.48, "learning_rate": 1.6559311862372477e-06, "loss": 0.0012, "step": 20870 }, { "epoch": 6.48, "learning_rate": 1.6555311062212446e-06, "loss": 0.001, "step": 20871 }, { "epoch": 6.48, "learning_rate": 1.655131026205241e-06, "loss": 0.0022, "step": 20872 }, { "epoch": 6.48, "learning_rate": 1.654730946189238e-06, "loss": 0.0055, "step": 20873 }, { "epoch": 6.48, "learning_rate": 1.654330866173235e-06, "loss": 0.001, "step": 20874 }, { "epoch": 6.48, "learning_rate": 1.6539307861572316e-06, "loss": 0.0034, "step": 20875 }, { "epoch": 6.48, "learning_rate": 1.6535307061412285e-06, "loss": 0.0057, "step": 20876 }, { "epoch": 6.48, "learning_rate": 1.6531306261252254e-06, "loss": 0.0242, "step": 20877 }, { "epoch": 6.48, "learning_rate": 1.6527305461092219e-06, "loss": 0.0016, "step": 20878 }, { "epoch": 6.48, "learning_rate": 1.6523304660932188e-06, "loss": 0.0012, "step": 20879 }, { "epoch": 6.48, "learning_rate": 1.6519303860772156e-06, "loss": 0.0021, "step": 20880 }, { "epoch": 6.48, "learning_rate": 1.6515303060612123e-06, "loss": 0.0012, "step": 20881 }, { "epoch": 6.48, "learning_rate": 1.6511302260452092e-06, "loss": 0.0089, "step": 20882 }, { "epoch": 6.48, "learning_rate": 1.6507301460292061e-06, "loss": 0.0023, "step": 20883 }, { "epoch": 6.48, "learning_rate": 1.6503300660132026e-06, "loss": 0.0017, "step": 20884 }, { "epoch": 6.48, "learning_rate": 1.6499299859971995e-06, "loss": 0.0116, "step": 20885 }, { "epoch": 6.48, "learning_rate": 1.6495299059811964e-06, "loss": 0.0047, "step": 20886 }, { "epoch": 6.48, "learning_rate": 1.649129825965193e-06, "loss": 0.0012, "step": 20887 }, { "epoch": 6.48, "learning_rate": 1.64872974594919e-06, "loss": 0.0009, "step": 20888 }, { "epoch": 6.48, "learning_rate": 1.6483296659331869e-06, "loss": 0.0009, "step": 20889 }, { "epoch": 6.48, "learning_rate": 1.6479295859171834e-06, "loss": 0.0011, "step": 20890 }, { "epoch": 6.48, "learning_rate": 1.6475295059011803e-06, "loss": 0.0014, "step": 20891 }, { "epoch": 6.48, "learning_rate": 1.6471294258851772e-06, "loss": 0.0031, "step": 20892 }, { "epoch": 6.48, "learning_rate": 1.646729345869174e-06, "loss": 0.0016, "step": 20893 }, { "epoch": 6.49, "learning_rate": 1.6463292658531707e-06, "loss": 0.0063, "step": 20894 }, { "epoch": 6.49, "learning_rate": 1.6459291858371676e-06, "loss": 0.0028, "step": 20895 }, { "epoch": 6.49, "learning_rate": 1.6455291058211645e-06, "loss": 0.0084, "step": 20896 }, { "epoch": 6.49, "learning_rate": 1.645129025805161e-06, "loss": 0.0025, "step": 20897 }, { "epoch": 6.49, "learning_rate": 1.644728945789158e-06, "loss": 0.0015, "step": 20898 }, { "epoch": 6.49, "learning_rate": 1.6443288657731548e-06, "loss": 0.0022, "step": 20899 }, { "epoch": 6.49, "learning_rate": 1.6439287857571515e-06, "loss": 0.0015, "step": 20900 }, { "epoch": 6.49, "learning_rate": 1.6435287057411484e-06, "loss": 0.0022, "step": 20901 }, { "epoch": 6.49, "learning_rate": 1.6431286257251453e-06, "loss": 0.0052, "step": 20902 }, { "epoch": 6.49, "learning_rate": 1.6427285457091418e-06, "loss": 0.0013, "step": 20903 }, { "epoch": 6.49, "learning_rate": 1.6423284656931387e-06, "loss": 0.0031, "step": 20904 }, { "epoch": 6.49, "learning_rate": 1.6419283856771356e-06, "loss": 0.0039, "step": 20905 }, { "epoch": 6.49, "learning_rate": 1.6415283056611322e-06, "loss": 0.0059, "step": 20906 }, { "epoch": 6.49, "learning_rate": 1.6411282256451291e-06, "loss": 0.0028, "step": 20907 }, { "epoch": 6.49, "learning_rate": 1.640728145629126e-06, "loss": 0.0014, "step": 20908 }, { "epoch": 6.49, "learning_rate": 1.6403280656131227e-06, "loss": 0.0023, "step": 20909 }, { "epoch": 6.49, "learning_rate": 1.6399279855971196e-06, "loss": 0.0015, "step": 20910 }, { "epoch": 6.49, "learning_rate": 1.6395279055811163e-06, "loss": 0.0016, "step": 20911 }, { "epoch": 6.49, "learning_rate": 1.6391278255651132e-06, "loss": 0.0048, "step": 20912 }, { "epoch": 6.49, "learning_rate": 1.63872774554911e-06, "loss": 0.0018, "step": 20913 }, { "epoch": 6.49, "learning_rate": 1.6383276655331068e-06, "loss": 0.0023, "step": 20914 }, { "epoch": 6.49, "learning_rate": 1.6379275855171037e-06, "loss": 0.0048, "step": 20915 }, { "epoch": 6.49, "learning_rate": 1.6375275055011004e-06, "loss": 0.0009, "step": 20916 }, { "epoch": 6.49, "learning_rate": 1.6371274254850973e-06, "loss": 0.0015, "step": 20917 }, { "epoch": 6.49, "learning_rate": 1.636727345469094e-06, "loss": 0.0017, "step": 20918 }, { "epoch": 6.49, "learning_rate": 1.6363272654530906e-06, "loss": 0.0004, "step": 20919 }, { "epoch": 6.49, "learning_rate": 1.6359271854370875e-06, "loss": 0.0022, "step": 20920 }, { "epoch": 6.49, "learning_rate": 1.6355271054210844e-06, "loss": 0.0009, "step": 20921 }, { "epoch": 6.49, "learning_rate": 1.6351270254050811e-06, "loss": 0.0006, "step": 20922 }, { "epoch": 6.49, "learning_rate": 1.634726945389078e-06, "loss": 0.0012, "step": 20923 }, { "epoch": 6.49, "learning_rate": 1.634326865373075e-06, "loss": 0.0035, "step": 20924 }, { "epoch": 6.49, "learning_rate": 1.6339267853570714e-06, "loss": 0.0034, "step": 20925 }, { "epoch": 6.49, "learning_rate": 1.6335267053410683e-06, "loss": 0.0471, "step": 20926 }, { "epoch": 6.5, "learning_rate": 1.6331266253250652e-06, "loss": 0.0046, "step": 20927 }, { "epoch": 6.5, "learning_rate": 1.6327265453090619e-06, "loss": 0.0021, "step": 20928 }, { "epoch": 6.5, "learning_rate": 1.6323264652930588e-06, "loss": 0.0079, "step": 20929 }, { "epoch": 6.5, "learning_rate": 1.6319263852770557e-06, "loss": 0.0076, "step": 20930 }, { "epoch": 6.5, "learning_rate": 1.6315263052610522e-06, "loss": 0.0011, "step": 20931 }, { "epoch": 6.5, "learning_rate": 1.631126225245049e-06, "loss": 0.002, "step": 20932 }, { "epoch": 6.5, "learning_rate": 1.630726145229046e-06, "loss": 0.0009, "step": 20933 }, { "epoch": 6.5, "learning_rate": 1.6303260652130428e-06, "loss": 0.0021, "step": 20934 }, { "epoch": 6.5, "learning_rate": 1.6299259851970395e-06, "loss": 0.0011, "step": 20935 }, { "epoch": 6.5, "learning_rate": 1.6295259051810364e-06, "loss": 0.0006, "step": 20936 }, { "epoch": 6.5, "learning_rate": 1.6291258251650333e-06, "loss": 0.0072, "step": 20937 }, { "epoch": 6.5, "learning_rate": 1.6287257451490298e-06, "loss": 0.0021, "step": 20938 }, { "epoch": 6.5, "learning_rate": 1.6283256651330267e-06, "loss": 0.002, "step": 20939 }, { "epoch": 6.5, "learning_rate": 1.6279255851170236e-06, "loss": 0.0012, "step": 20940 }, { "epoch": 6.5, "learning_rate": 1.6275255051010203e-06, "loss": 0.0028, "step": 20941 }, { "epoch": 6.5, "learning_rate": 1.6271254250850172e-06, "loss": 0.0017, "step": 20942 }, { "epoch": 6.5, "learning_rate": 1.626725345069014e-06, "loss": 0.0055, "step": 20943 }, { "epoch": 6.5, "learning_rate": 1.6263252650530106e-06, "loss": 0.0013, "step": 20944 }, { "epoch": 6.5, "learning_rate": 1.6259251850370075e-06, "loss": 0.0029, "step": 20945 }, { "epoch": 6.5, "learning_rate": 1.6255251050210044e-06, "loss": 0.0013, "step": 20946 }, { "epoch": 6.5, "learning_rate": 1.625125025005001e-06, "loss": 0.0021, "step": 20947 }, { "epoch": 6.5, "learning_rate": 1.624724944988998e-06, "loss": 0.0013, "step": 20948 }, { "epoch": 6.5, "learning_rate": 1.6243248649729948e-06, "loss": 0.0018, "step": 20949 }, { "epoch": 6.5, "learning_rate": 1.6239247849569913e-06, "loss": 0.001, "step": 20950 }, { "epoch": 6.5, "learning_rate": 1.6235247049409882e-06, "loss": 0.0012, "step": 20951 }, { "epoch": 6.5, "learning_rate": 1.6231246249249851e-06, "loss": 0.0011, "step": 20952 }, { "epoch": 6.5, "learning_rate": 1.6227245449089818e-06, "loss": 0.0332, "step": 20953 }, { "epoch": 6.5, "learning_rate": 1.6223244648929787e-06, "loss": 0.0016, "step": 20954 }, { "epoch": 6.5, "learning_rate": 1.6219243848769756e-06, "loss": 0.0022, "step": 20955 }, { "epoch": 6.5, "learning_rate": 1.6215243048609725e-06, "loss": 0.0011, "step": 20956 }, { "epoch": 6.5, "learning_rate": 1.621124224844969e-06, "loss": 0.0013, "step": 20957 }, { "epoch": 6.5, "learning_rate": 1.6207241448289659e-06, "loss": 0.0015, "step": 20958 }, { "epoch": 6.51, "learning_rate": 1.6203240648129628e-06, "loss": 0.0013, "step": 20959 }, { "epoch": 6.51, "learning_rate": 1.6199239847969594e-06, "loss": 0.0015, "step": 20960 }, { "epoch": 6.51, "learning_rate": 1.6195239047809563e-06, "loss": 0.0004, "step": 20961 }, { "epoch": 6.51, "learning_rate": 1.6191238247649532e-06, "loss": 0.0037, "step": 20962 }, { "epoch": 6.51, "learning_rate": 1.61872374474895e-06, "loss": 0.0072, "step": 20963 }, { "epoch": 6.51, "learning_rate": 1.6183236647329466e-06, "loss": 0.0562, "step": 20964 }, { "epoch": 6.51, "learning_rate": 1.6179235847169435e-06, "loss": 0.0015, "step": 20965 }, { "epoch": 6.51, "learning_rate": 1.6175235047009402e-06, "loss": 0.0017, "step": 20966 }, { "epoch": 6.51, "learning_rate": 1.617123424684937e-06, "loss": 0.0014, "step": 20967 }, { "epoch": 6.51, "learning_rate": 1.616723344668934e-06, "loss": 0.0329, "step": 20968 }, { "epoch": 6.51, "learning_rate": 1.6163232646529307e-06, "loss": 0.0029, "step": 20969 }, { "epoch": 6.51, "learning_rate": 1.6159231846369276e-06, "loss": 0.0024, "step": 20970 }, { "epoch": 6.51, "learning_rate": 1.6155231046209245e-06, "loss": 0.0031, "step": 20971 }, { "epoch": 6.51, "learning_rate": 1.615123024604921e-06, "loss": 0.0018, "step": 20972 }, { "epoch": 6.51, "learning_rate": 1.6147229445889178e-06, "loss": 0.0017, "step": 20973 }, { "epoch": 6.51, "learning_rate": 1.6143228645729147e-06, "loss": 0.0023, "step": 20974 }, { "epoch": 6.51, "learning_rate": 1.6139227845569116e-06, "loss": 0.0066, "step": 20975 }, { "epoch": 6.51, "learning_rate": 1.6135227045409083e-06, "loss": 0.0268, "step": 20976 }, { "epoch": 6.51, "learning_rate": 1.6131226245249052e-06, "loss": 0.0022, "step": 20977 }, { "epoch": 6.51, "learning_rate": 1.6127225445089021e-06, "loss": 0.0024, "step": 20978 }, { "epoch": 6.51, "learning_rate": 1.6123224644928986e-06, "loss": 0.0014, "step": 20979 }, { "epoch": 6.51, "learning_rate": 1.6119223844768955e-06, "loss": 0.0007, "step": 20980 }, { "epoch": 6.51, "learning_rate": 1.6115223044608924e-06, "loss": 0.0016, "step": 20981 }, { "epoch": 6.51, "learning_rate": 1.611122224444889e-06, "loss": 0.0038, "step": 20982 }, { "epoch": 6.51, "learning_rate": 1.610722144428886e-06, "loss": 0.0013, "step": 20983 }, { "epoch": 6.51, "learning_rate": 1.6103220644128829e-06, "loss": 0.0009, "step": 20984 }, { "epoch": 6.51, "learning_rate": 1.6099219843968794e-06, "loss": 0.0083, "step": 20985 }, { "epoch": 6.51, "learning_rate": 1.6095219043808763e-06, "loss": 0.0219, "step": 20986 }, { "epoch": 6.51, "learning_rate": 1.6091218243648732e-06, "loss": 0.0024, "step": 20987 }, { "epoch": 6.51, "learning_rate": 1.6087217443488698e-06, "loss": 0.0065, "step": 20988 }, { "epoch": 6.51, "learning_rate": 1.6083216643328667e-06, "loss": 0.0014, "step": 20989 }, { "epoch": 6.51, "learning_rate": 1.6079215843168636e-06, "loss": 0.002, "step": 20990 }, { "epoch": 6.52, "learning_rate": 1.6075215043008601e-06, "loss": 0.003, "step": 20991 }, { "epoch": 6.52, "learning_rate": 1.607121424284857e-06, "loss": 0.0011, "step": 20992 }, { "epoch": 6.52, "learning_rate": 1.606721344268854e-06, "loss": 0.0015, "step": 20993 }, { "epoch": 6.52, "learning_rate": 1.6063212642528506e-06, "loss": 0.0594, "step": 20994 }, { "epoch": 6.52, "learning_rate": 1.6059211842368475e-06, "loss": 0.0015, "step": 20995 }, { "epoch": 6.52, "learning_rate": 1.6055211042208444e-06, "loss": 0.002, "step": 20996 }, { "epoch": 6.52, "learning_rate": 1.6051210242048413e-06, "loss": 0.0028, "step": 20997 }, { "epoch": 6.52, "learning_rate": 1.6047209441888378e-06, "loss": 0.003, "step": 20998 }, { "epoch": 6.52, "learning_rate": 1.6043208641728347e-06, "loss": 0.0008, "step": 20999 }, { "epoch": 6.52, "learning_rate": 1.6039207841568316e-06, "loss": 0.0017, "step": 21000 }, { "epoch": 6.52, "learning_rate": 1.6035207041408282e-06, "loss": 0.005, "step": 21001 }, { "epoch": 6.52, "learning_rate": 1.6031206241248251e-06, "loss": 0.001, "step": 21002 }, { "epoch": 6.52, "learning_rate": 1.602720544108822e-06, "loss": 0.0082, "step": 21003 }, { "epoch": 6.52, "learning_rate": 1.6023204640928185e-06, "loss": 0.0006, "step": 21004 }, { "epoch": 6.52, "learning_rate": 1.6019203840768154e-06, "loss": 0.0066, "step": 21005 }, { "epoch": 6.52, "learning_rate": 1.6015203040608123e-06, "loss": 0.0021, "step": 21006 }, { "epoch": 6.52, "learning_rate": 1.601120224044809e-06, "loss": 0.0037, "step": 21007 }, { "epoch": 6.52, "learning_rate": 1.6007201440288059e-06, "loss": 0.0046, "step": 21008 }, { "epoch": 6.52, "learning_rate": 1.6003200640128028e-06, "loss": 0.0021, "step": 21009 }, { "epoch": 6.52, "learning_rate": 1.5999199839967995e-06, "loss": 0.0154, "step": 21010 }, { "epoch": 6.52, "learning_rate": 1.5995199039807962e-06, "loss": 0.0516, "step": 21011 }, { "epoch": 6.52, "learning_rate": 1.599119823964793e-06, "loss": 0.0012, "step": 21012 }, { "epoch": 6.52, "learning_rate": 1.5987197439487897e-06, "loss": 0.0013, "step": 21013 }, { "epoch": 6.52, "learning_rate": 1.5983196639327866e-06, "loss": 0.0081, "step": 21014 }, { "epoch": 6.52, "learning_rate": 1.5979195839167835e-06, "loss": 0.0119, "step": 21015 }, { "epoch": 6.52, "learning_rate": 1.5975195039007804e-06, "loss": 0.0069, "step": 21016 }, { "epoch": 6.52, "learning_rate": 1.5971194238847771e-06, "loss": 0.0099, "step": 21017 }, { "epoch": 6.52, "learning_rate": 1.5967193438687738e-06, "loss": 0.002, "step": 21018 }, { "epoch": 6.52, "learning_rate": 1.5963192638527707e-06, "loss": 0.0266, "step": 21019 }, { "epoch": 6.52, "learning_rate": 1.5959191838367674e-06, "loss": 0.0096, "step": 21020 }, { "epoch": 6.52, "learning_rate": 1.5955191038207643e-06, "loss": 0.0016, "step": 21021 }, { "epoch": 6.52, "learning_rate": 1.5951190238047612e-06, "loss": 0.0035, "step": 21022 }, { "epoch": 6.53, "learning_rate": 1.5947189437887579e-06, "loss": 0.0038, "step": 21023 }, { "epoch": 6.53, "learning_rate": 1.5943188637727548e-06, "loss": 0.0014, "step": 21024 }, { "epoch": 6.53, "learning_rate": 1.5939187837567517e-06, "loss": 0.002, "step": 21025 }, { "epoch": 6.53, "learning_rate": 1.5935187037407482e-06, "loss": 0.001, "step": 21026 }, { "epoch": 6.53, "learning_rate": 1.593118623724745e-06, "loss": 0.0018, "step": 21027 }, { "epoch": 6.53, "learning_rate": 1.592718543708742e-06, "loss": 0.0048, "step": 21028 }, { "epoch": 6.53, "learning_rate": 1.5923184636927386e-06, "loss": 0.0014, "step": 21029 }, { "epoch": 6.53, "learning_rate": 1.5919183836767355e-06, "loss": 0.0055, "step": 21030 }, { "epoch": 6.53, "learning_rate": 1.5915183036607324e-06, "loss": 0.0025, "step": 21031 }, { "epoch": 6.53, "learning_rate": 1.591118223644729e-06, "loss": 0.0175, "step": 21032 }, { "epoch": 6.53, "learning_rate": 1.5907181436287258e-06, "loss": 0.0026, "step": 21033 }, { "epoch": 6.53, "learning_rate": 1.5903180636127227e-06, "loss": 0.0052, "step": 21034 }, { "epoch": 6.53, "learning_rate": 1.5899179835967194e-06, "loss": 0.0048, "step": 21035 }, { "epoch": 6.53, "learning_rate": 1.5895179035807163e-06, "loss": 0.0009, "step": 21036 }, { "epoch": 6.53, "learning_rate": 1.5891178235647132e-06, "loss": 0.009, "step": 21037 }, { "epoch": 6.53, "learning_rate": 1.58871774354871e-06, "loss": 0.001, "step": 21038 }, { "epoch": 6.53, "learning_rate": 1.5883176635327066e-06, "loss": 0.0017, "step": 21039 }, { "epoch": 6.53, "learning_rate": 1.5879175835167035e-06, "loss": 0.0213, "step": 21040 }, { "epoch": 6.53, "learning_rate": 1.5875175035007004e-06, "loss": 0.0015, "step": 21041 }, { "epoch": 6.53, "learning_rate": 1.587117423484697e-06, "loss": 0.0013, "step": 21042 }, { "epoch": 6.53, "learning_rate": 1.586717343468694e-06, "loss": 0.0042, "step": 21043 }, { "epoch": 6.53, "learning_rate": 1.5863172634526908e-06, "loss": 0.0009, "step": 21044 }, { "epoch": 6.53, "learning_rate": 1.5859171834366873e-06, "loss": 0.0038, "step": 21045 }, { "epoch": 6.53, "learning_rate": 1.5855171034206842e-06, "loss": 0.0051, "step": 21046 }, { "epoch": 6.53, "learning_rate": 1.585117023404681e-06, "loss": 0.0008, "step": 21047 }, { "epoch": 6.53, "learning_rate": 1.5847169433886778e-06, "loss": 0.0046, "step": 21048 }, { "epoch": 6.53, "learning_rate": 1.5843168633726747e-06, "loss": 0.0012, "step": 21049 }, { "epoch": 6.53, "learning_rate": 1.5839167833566716e-06, "loss": 0.0239, "step": 21050 }, { "epoch": 6.53, "learning_rate": 1.583516703340668e-06, "loss": 0.0226, "step": 21051 }, { "epoch": 6.53, "learning_rate": 1.583116623324665e-06, "loss": 0.0071, "step": 21052 }, { "epoch": 6.53, "learning_rate": 1.5827165433086619e-06, "loss": 0.0027, "step": 21053 }, { "epoch": 6.53, "learning_rate": 1.5823164632926585e-06, "loss": 0.0009, "step": 21054 }, { "epoch": 6.54, "learning_rate": 1.5819163832766554e-06, "loss": 0.001, "step": 21055 }, { "epoch": 6.54, "learning_rate": 1.5815163032606523e-06, "loss": 0.001, "step": 21056 }, { "epoch": 6.54, "learning_rate": 1.5811162232446492e-06, "loss": 0.0016, "step": 21057 }, { "epoch": 6.54, "learning_rate": 1.5807161432286457e-06, "loss": 0.0015, "step": 21058 }, { "epoch": 6.54, "learning_rate": 1.5803160632126426e-06, "loss": 0.0018, "step": 21059 }, { "epoch": 6.54, "learning_rate": 1.5799159831966395e-06, "loss": 0.0011, "step": 21060 }, { "epoch": 6.54, "learning_rate": 1.5795159031806362e-06, "loss": 0.0024, "step": 21061 }, { "epoch": 6.54, "learning_rate": 1.579115823164633e-06, "loss": 0.0023, "step": 21062 }, { "epoch": 6.54, "learning_rate": 1.57871574314863e-06, "loss": 0.001, "step": 21063 }, { "epoch": 6.54, "learning_rate": 1.5783156631326267e-06, "loss": 0.0025, "step": 21064 }, { "epoch": 6.54, "learning_rate": 1.5779155831166234e-06, "loss": 0.0018, "step": 21065 }, { "epoch": 6.54, "learning_rate": 1.5775155031006203e-06, "loss": 0.0442, "step": 21066 }, { "epoch": 6.54, "learning_rate": 1.577115423084617e-06, "loss": 0.0025, "step": 21067 }, { "epoch": 6.54, "learning_rate": 1.5767153430686138e-06, "loss": 0.0387, "step": 21068 }, { "epoch": 6.54, "learning_rate": 1.5763152630526107e-06, "loss": 0.0019, "step": 21069 }, { "epoch": 6.54, "learning_rate": 1.5759151830366074e-06, "loss": 0.0017, "step": 21070 }, { "epoch": 6.54, "learning_rate": 1.5755151030206043e-06, "loss": 0.0013, "step": 21071 }, { "epoch": 6.54, "learning_rate": 1.575115023004601e-06, "loss": 0.023, "step": 21072 }, { "epoch": 6.54, "learning_rate": 1.5747149429885977e-06, "loss": 0.0011, "step": 21073 }, { "epoch": 6.54, "learning_rate": 1.5743148629725946e-06, "loss": 0.0014, "step": 21074 }, { "epoch": 6.54, "learning_rate": 1.5739147829565915e-06, "loss": 0.0012, "step": 21075 }, { "epoch": 6.54, "learning_rate": 1.5735147029405882e-06, "loss": 0.0015, "step": 21076 }, { "epoch": 6.54, "learning_rate": 1.573114622924585e-06, "loss": 0.0029, "step": 21077 }, { "epoch": 6.54, "learning_rate": 1.572714542908582e-06, "loss": 0.0011, "step": 21078 }, { "epoch": 6.54, "learning_rate": 1.5723144628925789e-06, "loss": 0.0158, "step": 21079 }, { "epoch": 6.54, "learning_rate": 1.5719143828765754e-06, "loss": 0.0034, "step": 21080 }, { "epoch": 6.54, "learning_rate": 1.5715143028605722e-06, "loss": 0.0011, "step": 21081 }, { "epoch": 6.54, "learning_rate": 1.5711142228445691e-06, "loss": 0.0008, "step": 21082 }, { "epoch": 6.54, "learning_rate": 1.5707141428285658e-06, "loss": 0.0013, "step": 21083 }, { "epoch": 6.54, "learning_rate": 1.5703140628125627e-06, "loss": 0.0011, "step": 21084 }, { "epoch": 6.54, "learning_rate": 1.5699139827965596e-06, "loss": 0.0015, "step": 21085 }, { "epoch": 6.54, "learning_rate": 1.569513902780556e-06, "loss": 0.0012, "step": 21086 }, { "epoch": 6.54, "learning_rate": 1.569113822764553e-06, "loss": 0.0024, "step": 21087 }, { "epoch": 6.55, "learning_rate": 1.56871374274855e-06, "loss": 0.0009, "step": 21088 }, { "epoch": 6.55, "learning_rate": 1.5683136627325466e-06, "loss": 0.0021, "step": 21089 }, { "epoch": 6.55, "learning_rate": 1.5679135827165435e-06, "loss": 0.002, "step": 21090 }, { "epoch": 6.55, "learning_rate": 1.5675135027005404e-06, "loss": 0.0019, "step": 21091 }, { "epoch": 6.55, "learning_rate": 1.5671134226845369e-06, "loss": 0.0012, "step": 21092 }, { "epoch": 6.55, "learning_rate": 1.5667133426685338e-06, "loss": 0.0007, "step": 21093 }, { "epoch": 6.55, "learning_rate": 1.5663132626525307e-06, "loss": 0.0028, "step": 21094 }, { "epoch": 6.55, "learning_rate": 1.5659131826365273e-06, "loss": 0.0028, "step": 21095 }, { "epoch": 6.55, "learning_rate": 1.5655131026205242e-06, "loss": 0.002, "step": 21096 }, { "epoch": 6.55, "learning_rate": 1.5651130226045211e-06, "loss": 0.0008, "step": 21097 }, { "epoch": 6.55, "learning_rate": 1.564712942588518e-06, "loss": 0.0014, "step": 21098 }, { "epoch": 6.55, "learning_rate": 1.5643128625725145e-06, "loss": 0.0018, "step": 21099 }, { "epoch": 6.55, "learning_rate": 1.5639127825565114e-06, "loss": 0.0011, "step": 21100 }, { "epoch": 6.55, "learning_rate": 1.5635127025405083e-06, "loss": 0.0021, "step": 21101 }, { "epoch": 6.55, "learning_rate": 1.563112622524505e-06, "loss": 0.0007, "step": 21102 }, { "epoch": 6.55, "learning_rate": 1.5627125425085019e-06, "loss": 0.0046, "step": 21103 }, { "epoch": 6.55, "learning_rate": 1.5623124624924988e-06, "loss": 0.002, "step": 21104 }, { "epoch": 6.55, "learning_rate": 1.5619123824764953e-06, "loss": 0.0017, "step": 21105 }, { "epoch": 6.55, "learning_rate": 1.5615123024604922e-06, "loss": 0.0035, "step": 21106 }, { "epoch": 6.55, "learning_rate": 1.561112222444489e-06, "loss": 0.019, "step": 21107 }, { "epoch": 6.55, "learning_rate": 1.5607121424284857e-06, "loss": 0.0008, "step": 21108 }, { "epoch": 6.55, "learning_rate": 1.5603120624124826e-06, "loss": 0.001, "step": 21109 }, { "epoch": 6.55, "learning_rate": 1.5599119823964795e-06, "loss": 0.0037, "step": 21110 }, { "epoch": 6.55, "learning_rate": 1.559511902380476e-06, "loss": 0.0114, "step": 21111 }, { "epoch": 6.55, "learning_rate": 1.559111822364473e-06, "loss": 0.0019, "step": 21112 }, { "epoch": 6.55, "learning_rate": 1.5587117423484698e-06, "loss": 0.0017, "step": 21113 }, { "epoch": 6.55, "learning_rate": 1.5583116623324665e-06, "loss": 0.0094, "step": 21114 }, { "epoch": 6.55, "learning_rate": 1.5579115823164634e-06, "loss": 0.0017, "step": 21115 }, { "epoch": 6.55, "learning_rate": 1.5575115023004603e-06, "loss": 0.0024, "step": 21116 }, { "epoch": 6.55, "learning_rate": 1.557111422284457e-06, "loss": 0.0013, "step": 21117 }, { "epoch": 6.55, "learning_rate": 1.5567113422684539e-06, "loss": 0.002, "step": 21118 }, { "epoch": 6.55, "learning_rate": 1.5563112622524506e-06, "loss": 0.001, "step": 21119 }, { "epoch": 6.56, "learning_rate": 1.5559111822364475e-06, "loss": 0.0016, "step": 21120 }, { "epoch": 6.56, "learning_rate": 1.5555111022204441e-06, "loss": 0.0121, "step": 21121 }, { "epoch": 6.56, "learning_rate": 1.555111022204441e-06, "loss": 0.0035, "step": 21122 }, { "epoch": 6.56, "learning_rate": 1.554710942188438e-06, "loss": 0.0029, "step": 21123 }, { "epoch": 6.56, "learning_rate": 1.5543108621724346e-06, "loss": 0.0013, "step": 21124 }, { "epoch": 6.56, "learning_rate": 1.5539107821564315e-06, "loss": 0.0012, "step": 21125 }, { "epoch": 6.56, "learning_rate": 1.5535107021404282e-06, "loss": 0.0022, "step": 21126 }, { "epoch": 6.56, "learning_rate": 1.553110622124425e-06, "loss": 0.0032, "step": 21127 }, { "epoch": 6.56, "learning_rate": 1.5527105421084218e-06, "loss": 0.0036, "step": 21128 }, { "epoch": 6.56, "learning_rate": 1.5523104620924187e-06, "loss": 0.0009, "step": 21129 }, { "epoch": 6.56, "learning_rate": 1.5519103820764154e-06, "loss": 0.0019, "step": 21130 }, { "epoch": 6.56, "learning_rate": 1.5515103020604123e-06, "loss": 0.0009, "step": 21131 }, { "epoch": 6.56, "learning_rate": 1.5511102220444092e-06, "loss": 0.0018, "step": 21132 }, { "epoch": 6.56, "learning_rate": 1.5507101420284057e-06, "loss": 0.0017, "step": 21133 }, { "epoch": 6.56, "learning_rate": 1.5503100620124026e-06, "loss": 0.0029, "step": 21134 }, { "epoch": 6.56, "learning_rate": 1.5499099819963994e-06, "loss": 0.0013, "step": 21135 }, { "epoch": 6.56, "learning_rate": 1.5495099019803961e-06, "loss": 0.0024, "step": 21136 }, { "epoch": 6.56, "learning_rate": 1.549109821964393e-06, "loss": 0.0018, "step": 21137 }, { "epoch": 6.56, "learning_rate": 1.54870974194839e-06, "loss": 0.0039, "step": 21138 }, { "epoch": 6.56, "learning_rate": 1.5483096619323864e-06, "loss": 0.0006, "step": 21139 }, { "epoch": 6.56, "learning_rate": 1.5479095819163833e-06, "loss": 0.0011, "step": 21140 }, { "epoch": 6.56, "learning_rate": 1.5475095019003802e-06, "loss": 0.0023, "step": 21141 }, { "epoch": 6.56, "learning_rate": 1.547109421884377e-06, "loss": 0.0043, "step": 21142 }, { "epoch": 6.56, "learning_rate": 1.5467093418683738e-06, "loss": 0.0014, "step": 21143 }, { "epoch": 6.56, "learning_rate": 1.5463092618523707e-06, "loss": 0.0019, "step": 21144 }, { "epoch": 6.56, "learning_rate": 1.5459091818363676e-06, "loss": 0.0021, "step": 21145 }, { "epoch": 6.56, "learning_rate": 1.545509101820364e-06, "loss": 0.0019, "step": 21146 }, { "epoch": 6.56, "learning_rate": 1.545109021804361e-06, "loss": 0.0013, "step": 21147 }, { "epoch": 6.56, "learning_rate": 1.5447089417883579e-06, "loss": 0.0015, "step": 21148 }, { "epoch": 6.56, "learning_rate": 1.5443088617723545e-06, "loss": 0.0065, "step": 21149 }, { "epoch": 6.56, "learning_rate": 1.5439087817563514e-06, "loss": 0.0224, "step": 21150 }, { "epoch": 6.56, "learning_rate": 1.5435087017403483e-06, "loss": 0.0047, "step": 21151 }, { "epoch": 6.57, "learning_rate": 1.5431086217243448e-06, "loss": 0.0017, "step": 21152 }, { "epoch": 6.57, "learning_rate": 1.5427085417083417e-06, "loss": 0.0016, "step": 21153 }, { "epoch": 6.57, "learning_rate": 1.5423084616923386e-06, "loss": 0.0013, "step": 21154 }, { "epoch": 6.57, "learning_rate": 1.5419083816763353e-06, "loss": 0.0032, "step": 21155 }, { "epoch": 6.57, "learning_rate": 1.5415083016603322e-06, "loss": 0.009, "step": 21156 }, { "epoch": 6.57, "learning_rate": 1.541108221644329e-06, "loss": 0.0045, "step": 21157 }, { "epoch": 6.57, "learning_rate": 1.5407081416283256e-06, "loss": 0.0009, "step": 21158 }, { "epoch": 6.57, "learning_rate": 1.5403080616123225e-06, "loss": 0.0022, "step": 21159 }, { "epoch": 6.57, "learning_rate": 1.5399079815963194e-06, "loss": 0.0007, "step": 21160 }, { "epoch": 6.57, "learning_rate": 1.5395079015803163e-06, "loss": 0.004, "step": 21161 }, { "epoch": 6.57, "learning_rate": 1.539107821564313e-06, "loss": 0.022, "step": 21162 }, { "epoch": 6.57, "learning_rate": 1.5387077415483098e-06, "loss": 0.0019, "step": 21163 }, { "epoch": 6.57, "learning_rate": 1.5383076615323067e-06, "loss": 0.0014, "step": 21164 }, { "epoch": 6.57, "learning_rate": 1.5379075815163032e-06, "loss": 0.0036, "step": 21165 }, { "epoch": 6.57, "learning_rate": 1.5375075015003001e-06, "loss": 0.0023, "step": 21166 }, { "epoch": 6.57, "learning_rate": 1.537107421484297e-06, "loss": 0.017, "step": 21167 }, { "epoch": 6.57, "learning_rate": 1.5367073414682937e-06, "loss": 0.0014, "step": 21168 }, { "epoch": 6.57, "learning_rate": 1.5363072614522906e-06, "loss": 0.0009, "step": 21169 }, { "epoch": 6.57, "learning_rate": 1.5359071814362875e-06, "loss": 0.0027, "step": 21170 }, { "epoch": 6.57, "learning_rate": 1.5355071014202842e-06, "loss": 0.0078, "step": 21171 }, { "epoch": 6.57, "learning_rate": 1.535107021404281e-06, "loss": 0.0034, "step": 21172 }, { "epoch": 6.57, "learning_rate": 1.5347069413882778e-06, "loss": 0.002, "step": 21173 }, { "epoch": 6.57, "learning_rate": 1.5343068613722744e-06, "loss": 0.0024, "step": 21174 }, { "epoch": 6.57, "learning_rate": 1.5339067813562713e-06, "loss": 0.0011, "step": 21175 }, { "epoch": 6.57, "learning_rate": 1.5335067013402682e-06, "loss": 0.0004, "step": 21176 }, { "epoch": 6.57, "learning_rate": 1.533106621324265e-06, "loss": 0.0013, "step": 21177 }, { "epoch": 6.57, "learning_rate": 1.5327065413082618e-06, "loss": 0.0063, "step": 21178 }, { "epoch": 6.57, "learning_rate": 1.5323064612922587e-06, "loss": 0.0008, "step": 21179 }, { "epoch": 6.57, "learning_rate": 1.5319063812762552e-06, "loss": 0.0008, "step": 21180 }, { "epoch": 6.57, "learning_rate": 1.531506301260252e-06, "loss": 0.0024, "step": 21181 }, { "epoch": 6.57, "learning_rate": 1.531106221244249e-06, "loss": 0.0009, "step": 21182 }, { "epoch": 6.57, "learning_rate": 1.530706141228246e-06, "loss": 0.0011, "step": 21183 }, { "epoch": 6.58, "learning_rate": 1.5303060612122426e-06, "loss": 0.004, "step": 21184 }, { "epoch": 6.58, "learning_rate": 1.5299059811962395e-06, "loss": 0.0043, "step": 21185 }, { "epoch": 6.58, "learning_rate": 1.5295059011802364e-06, "loss": 0.0037, "step": 21186 }, { "epoch": 6.58, "learning_rate": 1.5291058211642329e-06, "loss": 0.0023, "step": 21187 }, { "epoch": 6.58, "learning_rate": 1.5287057411482298e-06, "loss": 0.0008, "step": 21188 }, { "epoch": 6.58, "learning_rate": 1.5283056611322266e-06, "loss": 0.0098, "step": 21189 }, { "epoch": 6.58, "learning_rate": 1.5279055811162233e-06, "loss": 0.0099, "step": 21190 }, { "epoch": 6.58, "learning_rate": 1.5275055011002202e-06, "loss": 0.0012, "step": 21191 }, { "epoch": 6.58, "learning_rate": 1.5271054210842171e-06, "loss": 0.0118, "step": 21192 }, { "epoch": 6.58, "learning_rate": 1.5267053410682136e-06, "loss": 0.0167, "step": 21193 }, { "epoch": 6.58, "learning_rate": 1.5263052610522105e-06, "loss": 0.0021, "step": 21194 }, { "epoch": 6.58, "learning_rate": 1.5259051810362074e-06, "loss": 0.0014, "step": 21195 }, { "epoch": 6.58, "learning_rate": 1.525505101020204e-06, "loss": 0.0046, "step": 21196 }, { "epoch": 6.58, "learning_rate": 1.525105021004201e-06, "loss": 0.0033, "step": 21197 }, { "epoch": 6.58, "learning_rate": 1.5247049409881979e-06, "loss": 0.0007, "step": 21198 }, { "epoch": 6.58, "learning_rate": 1.5243048609721944e-06, "loss": 0.0016, "step": 21199 }, { "epoch": 6.58, "learning_rate": 1.5239047809561913e-06, "loss": 0.0028, "step": 21200 }, { "epoch": 6.58, "learning_rate": 1.5235047009401882e-06, "loss": 0.0011, "step": 21201 }, { "epoch": 6.58, "learning_rate": 1.523104620924185e-06, "loss": 0.0024, "step": 21202 }, { "epoch": 6.58, "learning_rate": 1.5227045409081817e-06, "loss": 0.0005, "step": 21203 }, { "epoch": 6.58, "learning_rate": 1.5223044608921786e-06, "loss": 0.0046, "step": 21204 }, { "epoch": 6.58, "learning_rate": 1.5219043808761755e-06, "loss": 0.0017, "step": 21205 }, { "epoch": 6.58, "learning_rate": 1.521504300860172e-06, "loss": 0.0018, "step": 21206 }, { "epoch": 6.58, "learning_rate": 1.521104220844169e-06, "loss": 0.0022, "step": 21207 }, { "epoch": 6.58, "learning_rate": 1.5207041408281658e-06, "loss": 0.0052, "step": 21208 }, { "epoch": 6.58, "learning_rate": 1.5203040608121625e-06, "loss": 0.0058, "step": 21209 }, { "epoch": 6.58, "learning_rate": 1.5199039807961594e-06, "loss": 0.0015, "step": 21210 }, { "epoch": 6.58, "learning_rate": 1.5195039007801563e-06, "loss": 0.0017, "step": 21211 }, { "epoch": 6.58, "learning_rate": 1.5191038207641528e-06, "loss": 0.0015, "step": 21212 }, { "epoch": 6.58, "learning_rate": 1.5187037407481497e-06, "loss": 0.0016, "step": 21213 }, { "epoch": 6.58, "learning_rate": 1.5183036607321466e-06, "loss": 0.0012, "step": 21214 }, { "epoch": 6.58, "learning_rate": 1.5179035807161432e-06, "loss": 0.0082, "step": 21215 }, { "epoch": 6.58, "learning_rate": 1.5175035007001401e-06, "loss": 0.0056, "step": 21216 }, { "epoch": 6.59, "learning_rate": 1.517103420684137e-06, "loss": 0.003, "step": 21217 }, { "epoch": 6.59, "learning_rate": 1.5167033406681337e-06, "loss": 0.0011, "step": 21218 }, { "epoch": 6.59, "learning_rate": 1.5163032606521304e-06, "loss": 0.0011, "step": 21219 }, { "epoch": 6.59, "learning_rate": 1.5159031806361273e-06, "loss": 0.0267, "step": 21220 }, { "epoch": 6.59, "learning_rate": 1.515503100620124e-06, "loss": 0.0024, "step": 21221 }, { "epoch": 6.59, "learning_rate": 1.515103020604121e-06, "loss": 0.0026, "step": 21222 }, { "epoch": 6.59, "learning_rate": 1.5147029405881178e-06, "loss": 0.0027, "step": 21223 }, { "epoch": 6.59, "learning_rate": 1.5143028605721147e-06, "loss": 0.002, "step": 21224 }, { "epoch": 6.59, "learning_rate": 1.5139027805561114e-06, "loss": 0.0011, "step": 21225 }, { "epoch": 6.59, "learning_rate": 1.5135027005401083e-06, "loss": 0.0052, "step": 21226 }, { "epoch": 6.59, "learning_rate": 1.513102620524105e-06, "loss": 0.0009, "step": 21227 }, { "epoch": 6.59, "learning_rate": 1.5127025405081016e-06, "loss": 0.0018, "step": 21228 }, { "epoch": 6.59, "learning_rate": 1.5123024604920985e-06, "loss": 0.0171, "step": 21229 }, { "epoch": 6.59, "learning_rate": 1.5119023804760954e-06, "loss": 0.0044, "step": 21230 }, { "epoch": 6.59, "learning_rate": 1.5115023004600921e-06, "loss": 0.001, "step": 21231 }, { "epoch": 6.59, "learning_rate": 1.511102220444089e-06, "loss": 0.0008, "step": 21232 }, { "epoch": 6.59, "learning_rate": 1.510702140428086e-06, "loss": 0.0021, "step": 21233 }, { "epoch": 6.59, "learning_rate": 1.5103020604120824e-06, "loss": 0.0012, "step": 21234 }, { "epoch": 6.59, "learning_rate": 1.5099019803960793e-06, "loss": 0.0019, "step": 21235 }, { "epoch": 6.59, "learning_rate": 1.5095019003800762e-06, "loss": 0.0081, "step": 21236 }, { "epoch": 6.59, "learning_rate": 1.5091018203640729e-06, "loss": 0.0045, "step": 21237 }, { "epoch": 6.59, "learning_rate": 1.5087017403480698e-06, "loss": 0.0004, "step": 21238 }, { "epoch": 6.59, "learning_rate": 1.5083016603320667e-06, "loss": 0.0016, "step": 21239 }, { "epoch": 6.59, "learning_rate": 1.5079015803160632e-06, "loss": 0.0013, "step": 21240 }, { "epoch": 6.59, "learning_rate": 1.50750150030006e-06, "loss": 0.0011, "step": 21241 }, { "epoch": 6.59, "learning_rate": 1.507101420284057e-06, "loss": 0.0016, "step": 21242 }, { "epoch": 6.59, "learning_rate": 1.5067013402680538e-06, "loss": 0.001, "step": 21243 }, { "epoch": 6.59, "learning_rate": 1.5063012602520505e-06, "loss": 0.0023, "step": 21244 }, { "epoch": 6.59, "learning_rate": 1.5059011802360474e-06, "loss": 0.0013, "step": 21245 }, { "epoch": 6.59, "learning_rate": 1.5055011002200443e-06, "loss": 0.0016, "step": 21246 }, { "epoch": 6.59, "learning_rate": 1.5051010202040408e-06, "loss": 0.0013, "step": 21247 }, { "epoch": 6.59, "learning_rate": 1.5047009401880377e-06, "loss": 0.0012, "step": 21248 }, { "epoch": 6.6, "learning_rate": 1.5043008601720346e-06, "loss": 0.0154, "step": 21249 }, { "epoch": 6.6, "learning_rate": 1.5039007801560313e-06, "loss": 0.0015, "step": 21250 }, { "epoch": 6.6, "learning_rate": 1.5035007001400282e-06, "loss": 0.0079, "step": 21251 }, { "epoch": 6.6, "learning_rate": 1.503100620124025e-06, "loss": 0.0025, "step": 21252 }, { "epoch": 6.6, "learning_rate": 1.5027005401080216e-06, "loss": 0.0011, "step": 21253 }, { "epoch": 6.6, "learning_rate": 1.5023004600920185e-06, "loss": 0.0043, "step": 21254 }, { "epoch": 6.6, "learning_rate": 1.5019003800760154e-06, "loss": 0.0483, "step": 21255 }, { "epoch": 6.6, "learning_rate": 1.501500300060012e-06, "loss": 0.002, "step": 21256 }, { "epoch": 6.6, "learning_rate": 1.501100220044009e-06, "loss": 0.0015, "step": 21257 }, { "epoch": 6.6, "learning_rate": 1.5007001400280058e-06, "loss": 0.0011, "step": 21258 }, { "epoch": 6.6, "learning_rate": 1.5003000600120023e-06, "loss": 0.0023, "step": 21259 }, { "epoch": 6.6, "learning_rate": 1.4998999799959992e-06, "loss": 0.0014, "step": 21260 }, { "epoch": 6.6, "learning_rate": 1.4994998999799961e-06, "loss": 0.0037, "step": 21261 }, { "epoch": 6.6, "learning_rate": 1.4990998199639928e-06, "loss": 0.0654, "step": 21262 }, { "epoch": 6.6, "learning_rate": 1.4986997399479897e-06, "loss": 0.0272, "step": 21263 }, { "epoch": 6.6, "learning_rate": 1.4982996599319866e-06, "loss": 0.0012, "step": 21264 }, { "epoch": 6.6, "learning_rate": 1.4978995799159835e-06, "loss": 0.0163, "step": 21265 }, { "epoch": 6.6, "learning_rate": 1.49749949989998e-06, "loss": 0.0034, "step": 21266 }, { "epoch": 6.6, "learning_rate": 1.4970994198839769e-06, "loss": 0.0007, "step": 21267 }, { "epoch": 6.6, "learning_rate": 1.4966993398679738e-06, "loss": 0.0013, "step": 21268 }, { "epoch": 6.6, "learning_rate": 1.4962992598519704e-06, "loss": 0.002, "step": 21269 }, { "epoch": 6.6, "learning_rate": 1.4958991798359673e-06, "loss": 0.0015, "step": 21270 }, { "epoch": 6.6, "learning_rate": 1.4954990998199642e-06, "loss": 0.0042, "step": 21271 }, { "epoch": 6.6, "learning_rate": 1.495099019803961e-06, "loss": 0.0021, "step": 21272 }, { "epoch": 6.6, "learning_rate": 1.4946989397879576e-06, "loss": 0.0007, "step": 21273 }, { "epoch": 6.6, "learning_rate": 1.4942988597719545e-06, "loss": 0.0008, "step": 21274 }, { "epoch": 6.6, "learning_rate": 1.4938987797559512e-06, "loss": 0.0008, "step": 21275 }, { "epoch": 6.6, "learning_rate": 1.493498699739948e-06, "loss": 0.0061, "step": 21276 }, { "epoch": 6.6, "learning_rate": 1.493098619723945e-06, "loss": 0.0009, "step": 21277 }, { "epoch": 6.6, "learning_rate": 1.4926985397079417e-06, "loss": 0.0405, "step": 21278 }, { "epoch": 6.6, "learning_rate": 1.4922984596919386e-06, "loss": 0.0019, "step": 21279 }, { "epoch": 6.6, "learning_rate": 1.4918983796759355e-06, "loss": 0.0047, "step": 21280 }, { "epoch": 6.61, "learning_rate": 1.491498299659932e-06, "loss": 0.0032, "step": 21281 }, { "epoch": 6.61, "learning_rate": 1.4910982196439288e-06, "loss": 0.0008, "step": 21282 }, { "epoch": 6.61, "learning_rate": 1.4906981396279257e-06, "loss": 0.0014, "step": 21283 }, { "epoch": 6.61, "learning_rate": 1.4902980596119226e-06, "loss": 0.0015, "step": 21284 }, { "epoch": 6.61, "learning_rate": 1.4898979795959193e-06, "loss": 0.0023, "step": 21285 }, { "epoch": 6.61, "learning_rate": 1.4894978995799162e-06, "loss": 0.0116, "step": 21286 }, { "epoch": 6.61, "learning_rate": 1.4890978195639131e-06, "loss": 0.0047, "step": 21287 }, { "epoch": 6.61, "learning_rate": 1.4886977395479096e-06, "loss": 0.0013, "step": 21288 }, { "epoch": 6.61, "learning_rate": 1.4882976595319065e-06, "loss": 0.0009, "step": 21289 }, { "epoch": 6.61, "learning_rate": 1.4878975795159034e-06, "loss": 0.0154, "step": 21290 }, { "epoch": 6.61, "learning_rate": 1.4874974994999e-06, "loss": 0.0022, "step": 21291 }, { "epoch": 6.61, "learning_rate": 1.487097419483897e-06, "loss": 0.0027, "step": 21292 }, { "epoch": 6.61, "learning_rate": 1.4866973394678939e-06, "loss": 0.0006, "step": 21293 }, { "epoch": 6.61, "learning_rate": 1.4862972594518904e-06, "loss": 0.0018, "step": 21294 }, { "epoch": 6.61, "learning_rate": 1.4858971794358873e-06, "loss": 0.001, "step": 21295 }, { "epoch": 6.61, "learning_rate": 1.4854970994198842e-06, "loss": 0.0029, "step": 21296 }, { "epoch": 6.61, "learning_rate": 1.4850970194038808e-06, "loss": 0.0049, "step": 21297 }, { "epoch": 6.61, "learning_rate": 1.4846969393878777e-06, "loss": 0.0013, "step": 21298 }, { "epoch": 6.61, "learning_rate": 1.4842968593718746e-06, "loss": 0.0032, "step": 21299 }, { "epoch": 6.61, "learning_rate": 1.4838967793558711e-06, "loss": 0.0019, "step": 21300 }, { "epoch": 6.61, "learning_rate": 1.483496699339868e-06, "loss": 0.0024, "step": 21301 }, { "epoch": 6.61, "learning_rate": 1.483096619323865e-06, "loss": 0.002, "step": 21302 }, { "epoch": 6.61, "learning_rate": 1.4826965393078616e-06, "loss": 0.0017, "step": 21303 }, { "epoch": 6.61, "learning_rate": 1.4822964592918585e-06, "loss": 0.0016, "step": 21304 }, { "epoch": 6.61, "learning_rate": 1.4818963792758554e-06, "loss": 0.0027, "step": 21305 }, { "epoch": 6.61, "learning_rate": 1.4814962992598523e-06, "loss": 0.0014, "step": 21306 }, { "epoch": 6.61, "learning_rate": 1.4810962192438488e-06, "loss": 0.0007, "step": 21307 }, { "epoch": 6.61, "learning_rate": 1.4806961392278457e-06, "loss": 0.0012, "step": 21308 }, { "epoch": 6.61, "learning_rate": 1.4802960592118426e-06, "loss": 0.0006, "step": 21309 }, { "epoch": 6.61, "learning_rate": 1.4798959791958392e-06, "loss": 0.0025, "step": 21310 }, { "epoch": 6.61, "learning_rate": 1.4794958991798361e-06, "loss": 0.0016, "step": 21311 }, { "epoch": 6.61, "learning_rate": 1.479095819163833e-06, "loss": 0.0012, "step": 21312 }, { "epoch": 6.62, "learning_rate": 1.4786957391478295e-06, "loss": 0.0013, "step": 21313 }, { "epoch": 6.62, "learning_rate": 1.4782956591318264e-06, "loss": 0.002, "step": 21314 }, { "epoch": 6.62, "learning_rate": 1.4778955791158233e-06, "loss": 0.007, "step": 21315 }, { "epoch": 6.62, "learning_rate": 1.47749549909982e-06, "loss": 0.0016, "step": 21316 }, { "epoch": 6.62, "learning_rate": 1.4770954190838169e-06, "loss": 0.0012, "step": 21317 }, { "epoch": 6.62, "learning_rate": 1.4766953390678138e-06, "loss": 0.0018, "step": 21318 }, { "epoch": 6.62, "learning_rate": 1.4762952590518103e-06, "loss": 0.0015, "step": 21319 }, { "epoch": 6.62, "learning_rate": 1.4758951790358072e-06, "loss": 0.0032, "step": 21320 }, { "epoch": 6.62, "learning_rate": 1.475495099019804e-06, "loss": 0.0016, "step": 21321 }, { "epoch": 6.62, "learning_rate": 1.4750950190038007e-06, "loss": 0.0013, "step": 21322 }, { "epoch": 6.62, "learning_rate": 1.4746949389877976e-06, "loss": 0.0014, "step": 21323 }, { "epoch": 6.62, "learning_rate": 1.4742948589717945e-06, "loss": 0.0025, "step": 21324 }, { "epoch": 6.62, "learning_rate": 1.4738947789557914e-06, "loss": 0.0032, "step": 21325 }, { "epoch": 6.62, "learning_rate": 1.4734946989397881e-06, "loss": 0.0019, "step": 21326 }, { "epoch": 6.62, "learning_rate": 1.4730946189237848e-06, "loss": 0.0033, "step": 21327 }, { "epoch": 6.62, "learning_rate": 1.4726945389077817e-06, "loss": 0.0016, "step": 21328 }, { "epoch": 6.62, "learning_rate": 1.4722944588917784e-06, "loss": 0.0006, "step": 21329 }, { "epoch": 6.62, "learning_rate": 1.4718943788757753e-06, "loss": 0.0392, "step": 21330 }, { "epoch": 6.62, "learning_rate": 1.4714942988597722e-06, "loss": 0.0023, "step": 21331 }, { "epoch": 6.62, "learning_rate": 1.4710942188437689e-06, "loss": 0.0053, "step": 21332 }, { "epoch": 6.62, "learning_rate": 1.4706941388277658e-06, "loss": 0.0044, "step": 21333 }, { "epoch": 6.62, "learning_rate": 1.4702940588117625e-06, "loss": 0.0063, "step": 21334 }, { "epoch": 6.62, "learning_rate": 1.4698939787957592e-06, "loss": 0.0031, "step": 21335 }, { "epoch": 6.62, "learning_rate": 1.469493898779756e-06, "loss": 0.0031, "step": 21336 }, { "epoch": 6.62, "learning_rate": 1.469093818763753e-06, "loss": 0.0073, "step": 21337 }, { "epoch": 6.62, "learning_rate": 1.4686937387477496e-06, "loss": 0.0087, "step": 21338 }, { "epoch": 6.62, "learning_rate": 1.4682936587317465e-06, "loss": 0.0014, "step": 21339 }, { "epoch": 6.62, "learning_rate": 1.4678935787157434e-06, "loss": 0.0007, "step": 21340 }, { "epoch": 6.62, "learning_rate": 1.46749349869974e-06, "loss": 0.0024, "step": 21341 }, { "epoch": 6.62, "learning_rate": 1.4670934186837368e-06, "loss": 0.0026, "step": 21342 }, { "epoch": 6.62, "learning_rate": 1.4666933386677337e-06, "loss": 0.0016, "step": 21343 }, { "epoch": 6.62, "learning_rate": 1.4662932586517304e-06, "loss": 0.0018, "step": 21344 }, { "epoch": 6.63, "learning_rate": 1.4658931786357273e-06, "loss": 0.0021, "step": 21345 }, { "epoch": 6.63, "learning_rate": 1.4654930986197242e-06, "loss": 0.0012, "step": 21346 }, { "epoch": 6.63, "learning_rate": 1.465093018603721e-06, "loss": 0.0011, "step": 21347 }, { "epoch": 6.63, "learning_rate": 1.4646929385877176e-06, "loss": 0.0017, "step": 21348 }, { "epoch": 6.63, "learning_rate": 1.4642928585717145e-06, "loss": 0.0008, "step": 21349 }, { "epoch": 6.63, "learning_rate": 1.4638927785557114e-06, "loss": 0.0026, "step": 21350 }, { "epoch": 6.63, "learning_rate": 1.463492698539708e-06, "loss": 0.0331, "step": 21351 }, { "epoch": 6.63, "learning_rate": 1.463092618523705e-06, "loss": 0.0011, "step": 21352 }, { "epoch": 6.63, "learning_rate": 1.4626925385077018e-06, "loss": 0.0026, "step": 21353 }, { "epoch": 6.63, "learning_rate": 1.4622924584916983e-06, "loss": 0.0006, "step": 21354 }, { "epoch": 6.63, "learning_rate": 1.4618923784756952e-06, "loss": 0.001, "step": 21355 }, { "epoch": 6.63, "learning_rate": 1.461492298459692e-06, "loss": 0.0011, "step": 21356 }, { "epoch": 6.63, "learning_rate": 1.4610922184436888e-06, "loss": 0.0015, "step": 21357 }, { "epoch": 6.63, "learning_rate": 1.4606921384276857e-06, "loss": 0.0037, "step": 21358 }, { "epoch": 6.63, "learning_rate": 1.4602920584116826e-06, "loss": 0.0035, "step": 21359 }, { "epoch": 6.63, "learning_rate": 1.459891978395679e-06, "loss": 0.0015, "step": 21360 }, { "epoch": 6.63, "learning_rate": 1.459491898379676e-06, "loss": 0.0016, "step": 21361 }, { "epoch": 6.63, "learning_rate": 1.4590918183636729e-06, "loss": 0.002, "step": 21362 }, { "epoch": 6.63, "learning_rate": 1.4586917383476695e-06, "loss": 0.0261, "step": 21363 }, { "epoch": 6.63, "learning_rate": 1.4582916583316664e-06, "loss": 0.0017, "step": 21364 }, { "epoch": 6.63, "learning_rate": 1.4578915783156633e-06, "loss": 0.0032, "step": 21365 }, { "epoch": 6.63, "learning_rate": 1.4574914982996598e-06, "loss": 0.0012, "step": 21366 }, { "epoch": 6.63, "learning_rate": 1.4570914182836567e-06, "loss": 0.0028, "step": 21367 }, { "epoch": 6.63, "learning_rate": 1.4566913382676536e-06, "loss": 0.0025, "step": 21368 }, { "epoch": 6.63, "learning_rate": 1.4562912582516505e-06, "loss": 0.0029, "step": 21369 }, { "epoch": 6.63, "learning_rate": 1.4558911782356472e-06, "loss": 0.0044, "step": 21370 }, { "epoch": 6.63, "learning_rate": 1.455491098219644e-06, "loss": 0.0041, "step": 21371 }, { "epoch": 6.63, "learning_rate": 1.455091018203641e-06, "loss": 0.0012, "step": 21372 }, { "epoch": 6.63, "learning_rate": 1.4546909381876375e-06, "loss": 0.0008, "step": 21373 }, { "epoch": 6.63, "learning_rate": 1.4542908581716344e-06, "loss": 0.0035, "step": 21374 }, { "epoch": 6.63, "learning_rate": 1.4538907781556313e-06, "loss": 0.0036, "step": 21375 }, { "epoch": 6.63, "learning_rate": 1.453490698139628e-06, "loss": 0.0081, "step": 21376 }, { "epoch": 6.63, "learning_rate": 1.4530906181236248e-06, "loss": 0.0008, "step": 21377 }, { "epoch": 6.64, "learning_rate": 1.4526905381076217e-06, "loss": 0.0014, "step": 21378 }, { "epoch": 6.64, "learning_rate": 1.4522904580916184e-06, "loss": 0.0006, "step": 21379 }, { "epoch": 6.64, "learning_rate": 1.4518903780756153e-06, "loss": 0.0018, "step": 21380 }, { "epoch": 6.64, "learning_rate": 1.451490298059612e-06, "loss": 0.0012, "step": 21381 }, { "epoch": 6.64, "learning_rate": 1.4510902180436087e-06, "loss": 0.0013, "step": 21382 }, { "epoch": 6.64, "learning_rate": 1.4506901380276056e-06, "loss": 0.0018, "step": 21383 }, { "epoch": 6.64, "learning_rate": 1.4502900580116025e-06, "loss": 0.0017, "step": 21384 }, { "epoch": 6.64, "learning_rate": 1.4498899779955992e-06, "loss": 0.0007, "step": 21385 }, { "epoch": 6.64, "learning_rate": 1.449489897979596e-06, "loss": 0.004, "step": 21386 }, { "epoch": 6.64, "learning_rate": 1.449089817963593e-06, "loss": 0.0013, "step": 21387 }, { "epoch": 6.64, "learning_rate": 1.4486897379475897e-06, "loss": 0.0305, "step": 21388 }, { "epoch": 6.64, "learning_rate": 1.4482896579315864e-06, "loss": 0.001, "step": 21389 }, { "epoch": 6.64, "learning_rate": 1.4478895779155832e-06, "loss": 0.0023, "step": 21390 }, { "epoch": 6.64, "learning_rate": 1.4474894978995801e-06, "loss": 0.0039, "step": 21391 }, { "epoch": 6.64, "learning_rate": 1.4470894178835768e-06, "loss": 0.0063, "step": 21392 }, { "epoch": 6.64, "learning_rate": 1.4466893378675737e-06, "loss": 0.0012, "step": 21393 }, { "epoch": 6.64, "learning_rate": 1.4462892578515706e-06, "loss": 0.0073, "step": 21394 }, { "epoch": 6.64, "learning_rate": 1.445889177835567e-06, "loss": 0.0021, "step": 21395 }, { "epoch": 6.64, "learning_rate": 1.445489097819564e-06, "loss": 0.0017, "step": 21396 }, { "epoch": 6.64, "learning_rate": 1.445089017803561e-06, "loss": 0.0138, "step": 21397 }, { "epoch": 6.64, "learning_rate": 1.4446889377875576e-06, "loss": 0.0047, "step": 21398 }, { "epoch": 6.64, "learning_rate": 1.4442888577715545e-06, "loss": 0.0028, "step": 21399 }, { "epoch": 6.64, "learning_rate": 1.4438887777555514e-06, "loss": 0.0034, "step": 21400 }, { "epoch": 6.64, "learning_rate": 1.4434886977395479e-06, "loss": 0.0313, "step": 21401 }, { "epoch": 6.64, "learning_rate": 1.4430886177235448e-06, "loss": 0.0018, "step": 21402 }, { "epoch": 6.64, "learning_rate": 1.4426885377075417e-06, "loss": 0.001, "step": 21403 }, { "epoch": 6.64, "learning_rate": 1.4422884576915383e-06, "loss": 0.0012, "step": 21404 }, { "epoch": 6.64, "learning_rate": 1.4418883776755352e-06, "loss": 0.0022, "step": 21405 }, { "epoch": 6.64, "learning_rate": 1.4414882976595321e-06, "loss": 0.005, "step": 21406 }, { "epoch": 6.64, "learning_rate": 1.4410882176435286e-06, "loss": 0.0018, "step": 21407 }, { "epoch": 6.64, "learning_rate": 1.4406881376275255e-06, "loss": 0.0731, "step": 21408 }, { "epoch": 6.64, "learning_rate": 1.4402880576115224e-06, "loss": 0.0019, "step": 21409 }, { "epoch": 6.65, "learning_rate": 1.4398879775955193e-06, "loss": 0.0078, "step": 21410 }, { "epoch": 6.65, "learning_rate": 1.439487897579516e-06, "loss": 0.0019, "step": 21411 }, { "epoch": 6.65, "learning_rate": 1.4390878175635129e-06, "loss": 0.0014, "step": 21412 }, { "epoch": 6.65, "learning_rate": 1.4386877375475098e-06, "loss": 0.0045, "step": 21413 }, { "epoch": 6.65, "learning_rate": 1.4382876575315063e-06, "loss": 0.0026, "step": 21414 }, { "epoch": 6.65, "learning_rate": 1.4378875775155032e-06, "loss": 0.0121, "step": 21415 }, { "epoch": 6.65, "learning_rate": 1.4374874974995e-06, "loss": 0.0007, "step": 21416 }, { "epoch": 6.65, "learning_rate": 1.4370874174834967e-06, "loss": 0.0016, "step": 21417 }, { "epoch": 6.65, "learning_rate": 1.4366873374674936e-06, "loss": 0.0026, "step": 21418 }, { "epoch": 6.65, "learning_rate": 1.4362872574514905e-06, "loss": 0.009, "step": 21419 }, { "epoch": 6.65, "learning_rate": 1.435887177435487e-06, "loss": 0.0008, "step": 21420 }, { "epoch": 6.65, "learning_rate": 1.435487097419484e-06, "loss": 0.0009, "step": 21421 }, { "epoch": 6.65, "learning_rate": 1.4350870174034808e-06, "loss": 0.0014, "step": 21422 }, { "epoch": 6.65, "learning_rate": 1.4346869373874775e-06, "loss": 0.0006, "step": 21423 }, { "epoch": 6.65, "learning_rate": 1.4342868573714744e-06, "loss": 0.0023, "step": 21424 }, { "epoch": 6.65, "learning_rate": 1.4338867773554713e-06, "loss": 0.0054, "step": 21425 }, { "epoch": 6.65, "learning_rate": 1.433486697339468e-06, "loss": 0.0017, "step": 21426 }, { "epoch": 6.65, "learning_rate": 1.4330866173234647e-06, "loss": 0.0011, "step": 21427 }, { "epoch": 6.65, "learning_rate": 1.4326865373074616e-06, "loss": 0.001, "step": 21428 }, { "epoch": 6.65, "learning_rate": 1.4322864572914585e-06, "loss": 0.0012, "step": 21429 }, { "epoch": 6.65, "learning_rate": 1.4318863772754551e-06, "loss": 0.0053, "step": 21430 }, { "epoch": 6.65, "learning_rate": 1.431486297259452e-06, "loss": 0.0007, "step": 21431 }, { "epoch": 6.65, "learning_rate": 1.431086217243449e-06, "loss": 0.0035, "step": 21432 }, { "epoch": 6.65, "learning_rate": 1.4306861372274456e-06, "loss": 0.0012, "step": 21433 }, { "epoch": 6.65, "learning_rate": 1.4302860572114425e-06, "loss": 0.0014, "step": 21434 }, { "epoch": 6.65, "learning_rate": 1.4298859771954392e-06, "loss": 0.0013, "step": 21435 }, { "epoch": 6.65, "learning_rate": 1.429485897179436e-06, "loss": 0.0019, "step": 21436 }, { "epoch": 6.65, "learning_rate": 1.4290858171634328e-06, "loss": 0.0019, "step": 21437 }, { "epoch": 6.65, "learning_rate": 1.4286857371474297e-06, "loss": 0.0013, "step": 21438 }, { "epoch": 6.65, "learning_rate": 1.4282856571314264e-06, "loss": 0.0016, "step": 21439 }, { "epoch": 6.65, "learning_rate": 1.4278855771154233e-06, "loss": 0.0008, "step": 21440 }, { "epoch": 6.65, "learning_rate": 1.4274854970994202e-06, "loss": 0.0038, "step": 21441 }, { "epoch": 6.66, "learning_rate": 1.4270854170834167e-06, "loss": 0.0008, "step": 21442 }, { "epoch": 6.66, "learning_rate": 1.4266853370674136e-06, "loss": 0.0033, "step": 21443 }, { "epoch": 6.66, "learning_rate": 1.4262852570514104e-06, "loss": 0.0106, "step": 21444 }, { "epoch": 6.66, "learning_rate": 1.4258851770354071e-06, "loss": 0.0081, "step": 21445 }, { "epoch": 6.66, "learning_rate": 1.425485097019404e-06, "loss": 0.0013, "step": 21446 }, { "epoch": 6.66, "learning_rate": 1.425085017003401e-06, "loss": 0.0058, "step": 21447 }, { "epoch": 6.66, "learning_rate": 1.4246849369873974e-06, "loss": 0.0009, "step": 21448 }, { "epoch": 6.66, "learning_rate": 1.4242848569713943e-06, "loss": 0.0039, "step": 21449 }, { "epoch": 6.66, "learning_rate": 1.4238847769553912e-06, "loss": 0.0035, "step": 21450 }, { "epoch": 6.66, "learning_rate": 1.423484696939388e-06, "loss": 0.002, "step": 21451 }, { "epoch": 6.66, "learning_rate": 1.4230846169233848e-06, "loss": 0.0017, "step": 21452 }, { "epoch": 6.66, "learning_rate": 1.4226845369073817e-06, "loss": 0.002, "step": 21453 }, { "epoch": 6.66, "learning_rate": 1.4222844568913786e-06, "loss": 0.001, "step": 21454 }, { "epoch": 6.66, "learning_rate": 1.421884376875375e-06, "loss": 0.0019, "step": 21455 }, { "epoch": 6.66, "learning_rate": 1.421484296859372e-06, "loss": 0.0013, "step": 21456 }, { "epoch": 6.66, "learning_rate": 1.4210842168433689e-06, "loss": 0.0026, "step": 21457 }, { "epoch": 6.66, "learning_rate": 1.4206841368273655e-06, "loss": 0.0013, "step": 21458 }, { "epoch": 6.66, "learning_rate": 1.4202840568113624e-06, "loss": 0.0021, "step": 21459 }, { "epoch": 6.66, "learning_rate": 1.4198839767953593e-06, "loss": 0.002, "step": 21460 }, { "epoch": 6.66, "learning_rate": 1.4194838967793558e-06, "loss": 0.0015, "step": 21461 }, { "epoch": 6.66, "learning_rate": 1.4190838167633527e-06, "loss": 0.0011, "step": 21462 }, { "epoch": 6.66, "learning_rate": 1.4186837367473496e-06, "loss": 0.0055, "step": 21463 }, { "epoch": 6.66, "learning_rate": 1.4182836567313463e-06, "loss": 0.023, "step": 21464 }, { "epoch": 6.66, "learning_rate": 1.4178835767153432e-06, "loss": 0.0013, "step": 21465 }, { "epoch": 6.66, "learning_rate": 1.41748349669934e-06, "loss": 0.0031, "step": 21466 }, { "epoch": 6.66, "learning_rate": 1.4170834166833366e-06, "loss": 0.0081, "step": 21467 }, { "epoch": 6.66, "learning_rate": 1.4166833366673335e-06, "loss": 0.0068, "step": 21468 }, { "epoch": 6.66, "learning_rate": 1.4162832566513304e-06, "loss": 0.0167, "step": 21469 }, { "epoch": 6.66, "learning_rate": 1.4158831766353273e-06, "loss": 0.0032, "step": 21470 }, { "epoch": 6.66, "learning_rate": 1.415483096619324e-06, "loss": 0.0026, "step": 21471 }, { "epoch": 6.66, "learning_rate": 1.4150830166033208e-06, "loss": 0.0027, "step": 21472 }, { "epoch": 6.66, "learning_rate": 1.4146829365873177e-06, "loss": 0.0027, "step": 21473 }, { "epoch": 6.67, "learning_rate": 1.4142828565713142e-06, "loss": 0.0071, "step": 21474 }, { "epoch": 6.67, "learning_rate": 1.4138827765553111e-06, "loss": 0.0255, "step": 21475 }, { "epoch": 6.67, "learning_rate": 1.413482696539308e-06, "loss": 0.0011, "step": 21476 }, { "epoch": 6.67, "learning_rate": 1.4130826165233047e-06, "loss": 0.0017, "step": 21477 }, { "epoch": 6.67, "learning_rate": 1.4126825365073016e-06, "loss": 0.0037, "step": 21478 }, { "epoch": 6.67, "learning_rate": 1.4122824564912985e-06, "loss": 0.0009, "step": 21479 }, { "epoch": 6.67, "learning_rate": 1.4118823764752952e-06, "loss": 0.0523, "step": 21480 }, { "epoch": 6.67, "learning_rate": 1.4114822964592919e-06, "loss": 0.011, "step": 21481 }, { "epoch": 6.67, "learning_rate": 1.4110822164432888e-06, "loss": 0.0019, "step": 21482 }, { "epoch": 6.67, "learning_rate": 1.4106821364272854e-06, "loss": 0.0175, "step": 21483 }, { "epoch": 6.67, "learning_rate": 1.4102820564112823e-06, "loss": 0.0012, "step": 21484 }, { "epoch": 6.67, "learning_rate": 1.4098819763952792e-06, "loss": 0.0037, "step": 21485 }, { "epoch": 6.67, "learning_rate": 1.409481896379276e-06, "loss": 0.0025, "step": 21486 }, { "epoch": 6.67, "learning_rate": 1.4090818163632728e-06, "loss": 0.0006, "step": 21487 }, { "epoch": 6.67, "learning_rate": 1.4086817363472697e-06, "loss": 0.0021, "step": 21488 }, { "epoch": 6.67, "learning_rate": 1.4082816563312662e-06, "loss": 0.0272, "step": 21489 }, { "epoch": 6.67, "learning_rate": 1.407881576315263e-06, "loss": 0.0033, "step": 21490 }, { "epoch": 6.67, "learning_rate": 1.40748149629926e-06, "loss": 0.0045, "step": 21491 }, { "epoch": 6.67, "learning_rate": 1.407081416283257e-06, "loss": 0.0009, "step": 21492 }, { "epoch": 6.67, "learning_rate": 1.4066813362672536e-06, "loss": 0.0158, "step": 21493 }, { "epoch": 6.67, "learning_rate": 1.4062812562512505e-06, "loss": 0.0013, "step": 21494 }, { "epoch": 6.67, "learning_rate": 1.4058811762352474e-06, "loss": 0.0017, "step": 21495 }, { "epoch": 6.67, "learning_rate": 1.4054810962192439e-06, "loss": 0.0013, "step": 21496 }, { "epoch": 6.67, "learning_rate": 1.4050810162032408e-06, "loss": 0.0011, "step": 21497 }, { "epoch": 6.67, "learning_rate": 1.4046809361872376e-06, "loss": 0.0007, "step": 21498 }, { "epoch": 6.67, "learning_rate": 1.4042808561712343e-06, "loss": 0.0008, "step": 21499 }, { "epoch": 6.67, "learning_rate": 1.4038807761552312e-06, "loss": 0.0026, "step": 21500 }, { "epoch": 6.67, "learning_rate": 1.4034806961392281e-06, "loss": 0.0007, "step": 21501 }, { "epoch": 6.67, "learning_rate": 1.4030806161232246e-06, "loss": 0.0015, "step": 21502 }, { "epoch": 6.67, "learning_rate": 1.4026805361072215e-06, "loss": 0.0007, "step": 21503 }, { "epoch": 6.67, "learning_rate": 1.4022804560912184e-06, "loss": 0.0052, "step": 21504 }, { "epoch": 6.67, "learning_rate": 1.401880376075215e-06, "loss": 0.0014, "step": 21505 }, { "epoch": 6.67, "learning_rate": 1.401480296059212e-06, "loss": 0.004, "step": 21506 }, { "epoch": 6.68, "learning_rate": 1.4010802160432089e-06, "loss": 0.0038, "step": 21507 }, { "epoch": 6.68, "learning_rate": 1.4006801360272054e-06, "loss": 0.0017, "step": 21508 }, { "epoch": 6.68, "learning_rate": 1.4002800560112023e-06, "loss": 0.001, "step": 21509 }, { "epoch": 6.68, "learning_rate": 1.3998799759951992e-06, "loss": 0.0014, "step": 21510 }, { "epoch": 6.68, "learning_rate": 1.399479895979196e-06, "loss": 0.0018, "step": 21511 }, { "epoch": 6.68, "learning_rate": 1.3990798159631927e-06, "loss": 0.0027, "step": 21512 }, { "epoch": 6.68, "learning_rate": 1.3986797359471896e-06, "loss": 0.0029, "step": 21513 }, { "epoch": 6.68, "learning_rate": 1.3982796559311865e-06, "loss": 0.0012, "step": 21514 }, { "epoch": 6.68, "learning_rate": 1.397879575915183e-06, "loss": 0.0015, "step": 21515 }, { "epoch": 6.68, "learning_rate": 1.39747949589918e-06, "loss": 0.0032, "step": 21516 }, { "epoch": 6.68, "learning_rate": 1.3970794158831768e-06, "loss": 0.0008, "step": 21517 }, { "epoch": 6.68, "learning_rate": 1.3966793358671735e-06, "loss": 0.0011, "step": 21518 }, { "epoch": 6.68, "learning_rate": 1.3962792558511704e-06, "loss": 0.0028, "step": 21519 }, { "epoch": 6.68, "learning_rate": 1.3958791758351673e-06, "loss": 0.0027, "step": 21520 }, { "epoch": 6.68, "learning_rate": 1.3954790958191638e-06, "loss": 0.0005, "step": 21521 }, { "epoch": 6.68, "learning_rate": 1.3950790158031607e-06, "loss": 0.0011, "step": 21522 }, { "epoch": 6.68, "learning_rate": 1.3946789357871576e-06, "loss": 0.0023, "step": 21523 }, { "epoch": 6.68, "learning_rate": 1.3942788557711542e-06, "loss": 0.0013, "step": 21524 }, { "epoch": 6.68, "learning_rate": 1.3938787757551511e-06, "loss": 0.0202, "step": 21525 }, { "epoch": 6.68, "learning_rate": 1.393478695739148e-06, "loss": 0.0331, "step": 21526 }, { "epoch": 6.68, "learning_rate": 1.3930786157231447e-06, "loss": 0.0013, "step": 21527 }, { "epoch": 6.68, "learning_rate": 1.3926785357071414e-06, "loss": 0.0033, "step": 21528 }, { "epoch": 6.68, "learning_rate": 1.3922784556911383e-06, "loss": 0.0029, "step": 21529 }, { "epoch": 6.68, "learning_rate": 1.391878375675135e-06, "loss": 0.0044, "step": 21530 }, { "epoch": 6.68, "learning_rate": 1.391478295659132e-06, "loss": 0.0043, "step": 21531 }, { "epoch": 6.68, "learning_rate": 1.3910782156431288e-06, "loss": 0.001, "step": 21532 }, { "epoch": 6.68, "learning_rate": 1.3906781356271257e-06, "loss": 0.0061, "step": 21533 }, { "epoch": 6.68, "learning_rate": 1.3902780556111224e-06, "loss": 0.0015, "step": 21534 }, { "epoch": 6.68, "learning_rate": 1.389877975595119e-06, "loss": 0.0005, "step": 21535 }, { "epoch": 6.68, "learning_rate": 1.389477895579116e-06, "loss": 0.0023, "step": 21536 }, { "epoch": 6.68, "learning_rate": 1.3890778155631126e-06, "loss": 0.0187, "step": 21537 }, { "epoch": 6.68, "learning_rate": 1.3886777355471095e-06, "loss": 0.001, "step": 21538 }, { "epoch": 6.69, "learning_rate": 1.3882776555311064e-06, "loss": 0.0013, "step": 21539 }, { "epoch": 6.69, "learning_rate": 1.3878775755151031e-06, "loss": 0.0008, "step": 21540 }, { "epoch": 6.69, "learning_rate": 1.3874774954991e-06, "loss": 0.0067, "step": 21541 }, { "epoch": 6.69, "learning_rate": 1.387077415483097e-06, "loss": 0.0009, "step": 21542 }, { "epoch": 6.69, "learning_rate": 1.3866773354670934e-06, "loss": 0.0163, "step": 21543 }, { "epoch": 6.69, "learning_rate": 1.3862772554510903e-06, "loss": 0.0024, "step": 21544 }, { "epoch": 6.69, "learning_rate": 1.3858771754350872e-06, "loss": 0.0018, "step": 21545 }, { "epoch": 6.69, "learning_rate": 1.3854770954190839e-06, "loss": 0.004, "step": 21546 }, { "epoch": 6.69, "learning_rate": 1.3850770154030808e-06, "loss": 0.0007, "step": 21547 }, { "epoch": 6.69, "learning_rate": 1.3846769353870777e-06, "loss": 0.0023, "step": 21548 }, { "epoch": 6.69, "learning_rate": 1.3842768553710742e-06, "loss": 0.001, "step": 21549 }, { "epoch": 6.69, "learning_rate": 1.383876775355071e-06, "loss": 0.0393, "step": 21550 }, { "epoch": 6.69, "learning_rate": 1.383476695339068e-06, "loss": 0.0026, "step": 21551 }, { "epoch": 6.69, "learning_rate": 1.3830766153230648e-06, "loss": 0.0297, "step": 21552 }, { "epoch": 6.69, "learning_rate": 1.3826765353070615e-06, "loss": 0.0024, "step": 21553 }, { "epoch": 6.69, "learning_rate": 1.3822764552910584e-06, "loss": 0.0029, "step": 21554 }, { "epoch": 6.69, "learning_rate": 1.3818763752750553e-06, "loss": 0.0019, "step": 21555 }, { "epoch": 6.69, "learning_rate": 1.3814762952590518e-06, "loss": 0.0036, "step": 21556 }, { "epoch": 6.69, "learning_rate": 1.3810762152430487e-06, "loss": 0.0033, "step": 21557 }, { "epoch": 6.69, "learning_rate": 1.3806761352270456e-06, "loss": 0.0069, "step": 21558 }, { "epoch": 6.69, "learning_rate": 1.3802760552110423e-06, "loss": 0.0012, "step": 21559 }, { "epoch": 6.69, "learning_rate": 1.3798759751950392e-06, "loss": 0.0147, "step": 21560 }, { "epoch": 6.69, "learning_rate": 1.379475895179036e-06, "loss": 0.0028, "step": 21561 }, { "epoch": 6.69, "learning_rate": 1.3790758151630326e-06, "loss": 0.0016, "step": 21562 }, { "epoch": 6.69, "learning_rate": 1.3786757351470295e-06, "loss": 0.0011, "step": 21563 }, { "epoch": 6.69, "learning_rate": 1.3782756551310264e-06, "loss": 0.0015, "step": 21564 }, { "epoch": 6.69, "learning_rate": 1.377875575115023e-06, "loss": 0.0005, "step": 21565 }, { "epoch": 6.69, "learning_rate": 1.37747549509902e-06, "loss": 0.0008, "step": 21566 }, { "epoch": 6.69, "learning_rate": 1.3770754150830168e-06, "loss": 0.004, "step": 21567 }, { "epoch": 6.69, "learning_rate": 1.3766753350670133e-06, "loss": 0.0008, "step": 21568 }, { "epoch": 6.69, "learning_rate": 1.3762752550510102e-06, "loss": 0.0055, "step": 21569 }, { "epoch": 6.69, "learning_rate": 1.3758751750350071e-06, "loss": 0.009, "step": 21570 }, { "epoch": 6.7, "learning_rate": 1.3754750950190038e-06, "loss": 0.0036, "step": 21571 }, { "epoch": 6.7, "learning_rate": 1.3750750150030007e-06, "loss": 0.0017, "step": 21572 }, { "epoch": 6.7, "learning_rate": 1.3746749349869976e-06, "loss": 0.0053, "step": 21573 }, { "epoch": 6.7, "learning_rate": 1.3742748549709945e-06, "loss": 0.0008, "step": 21574 }, { "epoch": 6.7, "learning_rate": 1.373874774954991e-06, "loss": 0.0092, "step": 21575 }, { "epoch": 6.7, "learning_rate": 1.3734746949389879e-06, "loss": 0.0011, "step": 21576 }, { "epoch": 6.7, "learning_rate": 1.3730746149229848e-06, "loss": 0.002, "step": 21577 }, { "epoch": 6.7, "learning_rate": 1.3726745349069814e-06, "loss": 0.0011, "step": 21578 }, { "epoch": 6.7, "learning_rate": 1.3722744548909783e-06, "loss": 0.0017, "step": 21579 }, { "epoch": 6.7, "learning_rate": 1.3718743748749752e-06, "loss": 0.0016, "step": 21580 }, { "epoch": 6.7, "learning_rate": 1.371474294858972e-06, "loss": 0.0059, "step": 21581 }, { "epoch": 6.7, "learning_rate": 1.3710742148429686e-06, "loss": 0.001, "step": 21582 }, { "epoch": 6.7, "learning_rate": 1.3706741348269655e-06, "loss": 0.0015, "step": 21583 }, { "epoch": 6.7, "learning_rate": 1.3702740548109622e-06, "loss": 0.0007, "step": 21584 }, { "epoch": 6.7, "learning_rate": 1.369873974794959e-06, "loss": 0.0198, "step": 21585 }, { "epoch": 6.7, "learning_rate": 1.369473894778956e-06, "loss": 0.0009, "step": 21586 }, { "epoch": 6.7, "learning_rate": 1.3690738147629527e-06, "loss": 0.001, "step": 21587 }, { "epoch": 6.7, "learning_rate": 1.3686737347469496e-06, "loss": 0.0017, "step": 21588 }, { "epoch": 6.7, "learning_rate": 1.3682736547309463e-06, "loss": 0.0013, "step": 21589 }, { "epoch": 6.7, "learning_rate": 1.367873574714943e-06, "loss": 0.0033, "step": 21590 }, { "epoch": 6.7, "learning_rate": 1.3674734946989398e-06, "loss": 0.0048, "step": 21591 }, { "epoch": 6.7, "learning_rate": 1.3670734146829367e-06, "loss": 0.001, "step": 21592 }, { "epoch": 6.7, "learning_rate": 1.3666733346669334e-06, "loss": 0.0006, "step": 21593 }, { "epoch": 6.7, "learning_rate": 1.3662732546509303e-06, "loss": 0.0033, "step": 21594 }, { "epoch": 6.7, "learning_rate": 1.3658731746349272e-06, "loss": 0.0025, "step": 21595 }, { "epoch": 6.7, "learning_rate": 1.3654730946189241e-06, "loss": 0.0013, "step": 21596 }, { "epoch": 6.7, "learning_rate": 1.3650730146029206e-06, "loss": 0.0014, "step": 21597 }, { "epoch": 6.7, "learning_rate": 1.3646729345869175e-06, "loss": 0.0095, "step": 21598 }, { "epoch": 6.7, "learning_rate": 1.3642728545709144e-06, "loss": 0.0051, "step": 21599 }, { "epoch": 6.7, "learning_rate": 1.363872774554911e-06, "loss": 0.0006, "step": 21600 }, { "epoch": 6.7, "learning_rate": 1.363472694538908e-06, "loss": 0.0089, "step": 21601 }, { "epoch": 6.7, "learning_rate": 1.3630726145229049e-06, "loss": 0.0023, "step": 21602 }, { "epoch": 6.71, "learning_rate": 1.3626725345069014e-06, "loss": 0.0011, "step": 21603 }, { "epoch": 6.71, "learning_rate": 1.3622724544908983e-06, "loss": 0.0026, "step": 21604 }, { "epoch": 6.71, "learning_rate": 1.3618723744748952e-06, "loss": 0.0335, "step": 21605 }, { "epoch": 6.71, "learning_rate": 1.3614722944588918e-06, "loss": 0.0012, "step": 21606 }, { "epoch": 6.71, "learning_rate": 1.3610722144428887e-06, "loss": 0.0015, "step": 21607 }, { "epoch": 6.71, "learning_rate": 1.3606721344268856e-06, "loss": 0.0009, "step": 21608 }, { "epoch": 6.71, "learning_rate": 1.3602720544108821e-06, "loss": 0.0008, "step": 21609 }, { "epoch": 6.71, "learning_rate": 1.359871974394879e-06, "loss": 0.0015, "step": 21610 }, { "epoch": 6.71, "learning_rate": 1.359471894378876e-06, "loss": 0.002, "step": 21611 }, { "epoch": 6.71, "learning_rate": 1.3590718143628726e-06, "loss": 0.0035, "step": 21612 }, { "epoch": 6.71, "learning_rate": 1.3586717343468695e-06, "loss": 0.0045, "step": 21613 }, { "epoch": 6.71, "learning_rate": 1.3582716543308664e-06, "loss": 0.0249, "step": 21614 }, { "epoch": 6.71, "learning_rate": 1.3578715743148633e-06, "loss": 0.035, "step": 21615 }, { "epoch": 6.71, "learning_rate": 1.3574714942988598e-06, "loss": 0.003, "step": 21616 }, { "epoch": 6.71, "learning_rate": 1.3570714142828567e-06, "loss": 0.0034, "step": 21617 }, { "epoch": 6.71, "learning_rate": 1.3566713342668536e-06, "loss": 0.0291, "step": 21618 }, { "epoch": 6.71, "learning_rate": 1.3562712542508502e-06, "loss": 0.0011, "step": 21619 }, { "epoch": 6.71, "learning_rate": 1.3558711742348471e-06, "loss": 0.0023, "step": 21620 }, { "epoch": 6.71, "learning_rate": 1.355471094218844e-06, "loss": 0.039, "step": 21621 }, { "epoch": 6.71, "learning_rate": 1.3550710142028405e-06, "loss": 0.0246, "step": 21622 }, { "epoch": 6.71, "learning_rate": 1.3546709341868374e-06, "loss": 0.0028, "step": 21623 }, { "epoch": 6.71, "learning_rate": 1.3542708541708343e-06, "loss": 0.0017, "step": 21624 }, { "epoch": 6.71, "learning_rate": 1.353870774154831e-06, "loss": 0.0015, "step": 21625 }, { "epoch": 6.71, "learning_rate": 1.3534706941388279e-06, "loss": 0.0042, "step": 21626 }, { "epoch": 6.71, "learning_rate": 1.3530706141228248e-06, "loss": 0.0031, "step": 21627 }, { "epoch": 6.71, "learning_rate": 1.3526705341068213e-06, "loss": 0.0011, "step": 21628 }, { "epoch": 6.71, "learning_rate": 1.3522704540908182e-06, "loss": 0.0013, "step": 21629 }, { "epoch": 6.71, "learning_rate": 1.351870374074815e-06, "loss": 0.0026, "step": 21630 }, { "epoch": 6.71, "learning_rate": 1.3514702940588117e-06, "loss": 0.0025, "step": 21631 }, { "epoch": 6.71, "learning_rate": 1.3510702140428086e-06, "loss": 0.0013, "step": 21632 }, { "epoch": 6.71, "learning_rate": 1.3506701340268055e-06, "loss": 0.002, "step": 21633 }, { "epoch": 6.71, "learning_rate": 1.3502700540108022e-06, "loss": 0.0021, "step": 21634 }, { "epoch": 6.72, "learning_rate": 1.3498699739947991e-06, "loss": 0.001, "step": 21635 }, { "epoch": 6.72, "learning_rate": 1.3494698939787958e-06, "loss": 0.0015, "step": 21636 }, { "epoch": 6.72, "learning_rate": 1.3490698139627927e-06, "loss": 0.0008, "step": 21637 }, { "epoch": 6.72, "learning_rate": 1.3486697339467894e-06, "loss": 0.0024, "step": 21638 }, { "epoch": 6.72, "learning_rate": 1.3482696539307863e-06, "loss": 0.0074, "step": 21639 }, { "epoch": 6.72, "learning_rate": 1.3478695739147832e-06, "loss": 0.0018, "step": 21640 }, { "epoch": 6.72, "learning_rate": 1.3474694938987799e-06, "loss": 0.0012, "step": 21641 }, { "epoch": 6.72, "learning_rate": 1.3470694138827768e-06, "loss": 0.0013, "step": 21642 }, { "epoch": 6.72, "learning_rate": 1.3466693338667735e-06, "loss": 0.0066, "step": 21643 }, { "epoch": 6.72, "learning_rate": 1.3462692538507702e-06, "loss": 0.0021, "step": 21644 }, { "epoch": 6.72, "learning_rate": 1.345869173834767e-06, "loss": 0.0009, "step": 21645 }, { "epoch": 6.72, "learning_rate": 1.345469093818764e-06, "loss": 0.0009, "step": 21646 }, { "epoch": 6.72, "learning_rate": 1.3450690138027606e-06, "loss": 0.001, "step": 21647 }, { "epoch": 6.72, "learning_rate": 1.3446689337867575e-06, "loss": 0.0044, "step": 21648 }, { "epoch": 6.72, "learning_rate": 1.3442688537707544e-06, "loss": 0.0013, "step": 21649 }, { "epoch": 6.72, "learning_rate": 1.343868773754751e-06, "loss": 0.0013, "step": 21650 }, { "epoch": 6.72, "learning_rate": 1.3434686937387478e-06, "loss": 0.0044, "step": 21651 }, { "epoch": 6.72, "learning_rate": 1.3430686137227447e-06, "loss": 0.0421, "step": 21652 }, { "epoch": 6.72, "learning_rate": 1.3426685337067414e-06, "loss": 0.0018, "step": 21653 }, { "epoch": 6.72, "learning_rate": 1.3422684536907383e-06, "loss": 0.0009, "step": 21654 }, { "epoch": 6.72, "learning_rate": 1.3418683736747352e-06, "loss": 0.0031, "step": 21655 }, { "epoch": 6.72, "learning_rate": 1.341468293658732e-06, "loss": 0.0021, "step": 21656 }, { "epoch": 6.72, "learning_rate": 1.3410682136427286e-06, "loss": 0.0136, "step": 21657 }, { "epoch": 6.72, "learning_rate": 1.3406681336267255e-06, "loss": 0.0026, "step": 21658 }, { "epoch": 6.72, "learning_rate": 1.3402680536107224e-06, "loss": 0.0016, "step": 21659 }, { "epoch": 6.72, "learning_rate": 1.339867973594719e-06, "loss": 0.0161, "step": 21660 }, { "epoch": 6.72, "learning_rate": 1.339467893578716e-06, "loss": 0.0015, "step": 21661 }, { "epoch": 6.72, "learning_rate": 1.3390678135627128e-06, "loss": 0.0036, "step": 21662 }, { "epoch": 6.72, "learning_rate": 1.3386677335467093e-06, "loss": 0.001, "step": 21663 }, { "epoch": 6.72, "learning_rate": 1.3382676535307062e-06, "loss": 0.001, "step": 21664 }, { "epoch": 6.72, "learning_rate": 1.337867573514703e-06, "loss": 0.0099, "step": 21665 }, { "epoch": 6.72, "learning_rate": 1.3374674934986998e-06, "loss": 0.0103, "step": 21666 }, { "epoch": 6.72, "learning_rate": 1.3370674134826967e-06, "loss": 0.002, "step": 21667 }, { "epoch": 6.73, "learning_rate": 1.3366673334666936e-06, "loss": 0.0085, "step": 21668 }, { "epoch": 6.73, "learning_rate": 1.33626725345069e-06, "loss": 0.0147, "step": 21669 }, { "epoch": 6.73, "learning_rate": 1.335867173434687e-06, "loss": 0.0014, "step": 21670 }, { "epoch": 6.73, "learning_rate": 1.3354670934186839e-06, "loss": 0.0023, "step": 21671 }, { "epoch": 6.73, "learning_rate": 1.3350670134026805e-06, "loss": 0.0021, "step": 21672 }, { "epoch": 6.73, "learning_rate": 1.3346669333866774e-06, "loss": 0.0009, "step": 21673 }, { "epoch": 6.73, "learning_rate": 1.3342668533706743e-06, "loss": 0.0006, "step": 21674 }, { "epoch": 6.73, "learning_rate": 1.3338667733546708e-06, "loss": 0.0066, "step": 21675 }, { "epoch": 6.73, "learning_rate": 1.3334666933386677e-06, "loss": 0.0014, "step": 21676 }, { "epoch": 6.73, "learning_rate": 1.3330666133226646e-06, "loss": 0.001, "step": 21677 }, { "epoch": 6.73, "learning_rate": 1.3326665333066615e-06, "loss": 0.0037, "step": 21678 }, { "epoch": 6.73, "learning_rate": 1.3322664532906582e-06, "loss": 0.0025, "step": 21679 }, { "epoch": 6.73, "learning_rate": 1.331866373274655e-06, "loss": 0.0011, "step": 21680 }, { "epoch": 6.73, "learning_rate": 1.331466293258652e-06, "loss": 0.001, "step": 21681 }, { "epoch": 6.73, "learning_rate": 1.3310662132426485e-06, "loss": 0.0029, "step": 21682 }, { "epoch": 6.73, "learning_rate": 1.3306661332266454e-06, "loss": 0.0029, "step": 21683 }, { "epoch": 6.73, "learning_rate": 1.3302660532106423e-06, "loss": 0.0048, "step": 21684 }, { "epoch": 6.73, "learning_rate": 1.329865973194639e-06, "loss": 0.0037, "step": 21685 }, { "epoch": 6.73, "learning_rate": 1.3294658931786358e-06, "loss": 0.003, "step": 21686 }, { "epoch": 6.73, "learning_rate": 1.3290658131626327e-06, "loss": 0.0028, "step": 21687 }, { "epoch": 6.73, "learning_rate": 1.3286657331466294e-06, "loss": 0.0012, "step": 21688 }, { "epoch": 6.73, "learning_rate": 1.3282656531306261e-06, "loss": 0.0011, "step": 21689 }, { "epoch": 6.73, "learning_rate": 1.327865573114623e-06, "loss": 0.0021, "step": 21690 }, { "epoch": 6.73, "learning_rate": 1.3274654930986197e-06, "loss": 0.0011, "step": 21691 }, { "epoch": 6.73, "learning_rate": 1.3270654130826166e-06, "loss": 0.0236, "step": 21692 }, { "epoch": 6.73, "learning_rate": 1.3266653330666135e-06, "loss": 0.0238, "step": 21693 }, { "epoch": 6.73, "learning_rate": 1.3262652530506102e-06, "loss": 0.0011, "step": 21694 }, { "epoch": 6.73, "learning_rate": 1.325865173034607e-06, "loss": 0.0038, "step": 21695 }, { "epoch": 6.73, "learning_rate": 1.325465093018604e-06, "loss": 0.0054, "step": 21696 }, { "epoch": 6.73, "learning_rate": 1.3250650130026007e-06, "loss": 0.0006, "step": 21697 }, { "epoch": 6.73, "learning_rate": 1.3246649329865974e-06, "loss": 0.0019, "step": 21698 }, { "epoch": 6.73, "learning_rate": 1.3242648529705942e-06, "loss": 0.0042, "step": 21699 }, { "epoch": 6.74, "learning_rate": 1.3238647729545911e-06, "loss": 0.0017, "step": 21700 }, { "epoch": 6.74, "learning_rate": 1.3234646929385878e-06, "loss": 0.0016, "step": 21701 }, { "epoch": 6.74, "learning_rate": 1.3230646129225847e-06, "loss": 0.0053, "step": 21702 }, { "epoch": 6.74, "learning_rate": 1.3226645329065816e-06, "loss": 0.0015, "step": 21703 }, { "epoch": 6.74, "learning_rate": 1.322264452890578e-06, "loss": 0.0005, "step": 21704 }, { "epoch": 6.74, "learning_rate": 1.321864372874575e-06, "loss": 0.0013, "step": 21705 }, { "epoch": 6.74, "learning_rate": 1.321464292858572e-06, "loss": 0.0786, "step": 21706 }, { "epoch": 6.74, "learning_rate": 1.3210642128425686e-06, "loss": 0.0008, "step": 21707 }, { "epoch": 6.74, "learning_rate": 1.3206641328265655e-06, "loss": 0.0021, "step": 21708 }, { "epoch": 6.74, "learning_rate": 1.3202640528105624e-06, "loss": 0.0013, "step": 21709 }, { "epoch": 6.74, "learning_rate": 1.3198639727945589e-06, "loss": 0.0018, "step": 21710 }, { "epoch": 6.74, "learning_rate": 1.3194638927785558e-06, "loss": 0.0221, "step": 21711 }, { "epoch": 6.74, "learning_rate": 1.3190638127625527e-06, "loss": 0.0016, "step": 21712 }, { "epoch": 6.74, "learning_rate": 1.3186637327465493e-06, "loss": 0.0017, "step": 21713 }, { "epoch": 6.74, "learning_rate": 1.3182636527305462e-06, "loss": 0.0009, "step": 21714 }, { "epoch": 6.74, "learning_rate": 1.3178635727145431e-06, "loss": 0.0054, "step": 21715 }, { "epoch": 6.74, "learning_rate": 1.3174634926985396e-06, "loss": 0.0019, "step": 21716 }, { "epoch": 6.74, "learning_rate": 1.3170634126825365e-06, "loss": 0.0015, "step": 21717 }, { "epoch": 6.74, "learning_rate": 1.3166633326665334e-06, "loss": 0.0022, "step": 21718 }, { "epoch": 6.74, "learning_rate": 1.3162632526505303e-06, "loss": 0.0015, "step": 21719 }, { "epoch": 6.74, "learning_rate": 1.315863172634527e-06, "loss": 0.001, "step": 21720 }, { "epoch": 6.74, "learning_rate": 1.3154630926185239e-06, "loss": 0.0008, "step": 21721 }, { "epoch": 6.74, "learning_rate": 1.3150630126025208e-06, "loss": 0.0016, "step": 21722 }, { "epoch": 6.74, "learning_rate": 1.3146629325865173e-06, "loss": 0.0026, "step": 21723 }, { "epoch": 6.74, "learning_rate": 1.3142628525705142e-06, "loss": 0.0017, "step": 21724 }, { "epoch": 6.74, "learning_rate": 1.313862772554511e-06, "loss": 0.0038, "step": 21725 }, { "epoch": 6.74, "learning_rate": 1.3134626925385077e-06, "loss": 0.0158, "step": 21726 }, { "epoch": 6.74, "learning_rate": 1.3130626125225046e-06, "loss": 0.0018, "step": 21727 }, { "epoch": 6.74, "learning_rate": 1.3126625325065015e-06, "loss": 0.0382, "step": 21728 }, { "epoch": 6.74, "learning_rate": 1.312262452490498e-06, "loss": 0.0005, "step": 21729 }, { "epoch": 6.74, "learning_rate": 1.311862372474495e-06, "loss": 0.0006, "step": 21730 }, { "epoch": 6.74, "learning_rate": 1.3114622924584918e-06, "loss": 0.0013, "step": 21731 }, { "epoch": 6.75, "learning_rate": 1.3110622124424885e-06, "loss": 0.0008, "step": 21732 }, { "epoch": 6.75, "learning_rate": 1.3106621324264854e-06, "loss": 0.0028, "step": 21733 }, { "epoch": 6.75, "learning_rate": 1.3102620524104823e-06, "loss": 0.0014, "step": 21734 }, { "epoch": 6.75, "learning_rate": 1.309861972394479e-06, "loss": 0.0029, "step": 21735 }, { "epoch": 6.75, "learning_rate": 1.3094618923784757e-06, "loss": 0.0199, "step": 21736 }, { "epoch": 6.75, "learning_rate": 1.3090618123624726e-06, "loss": 0.0531, "step": 21737 }, { "epoch": 6.75, "learning_rate": 1.3086617323464695e-06, "loss": 0.0015, "step": 21738 }, { "epoch": 6.75, "learning_rate": 1.3082616523304661e-06, "loss": 0.0434, "step": 21739 }, { "epoch": 6.75, "learning_rate": 1.307861572314463e-06, "loss": 0.0028, "step": 21740 }, { "epoch": 6.75, "learning_rate": 1.30746149229846e-06, "loss": 0.0042, "step": 21741 }, { "epoch": 6.75, "learning_rate": 1.3070614122824566e-06, "loss": 0.001, "step": 21742 }, { "epoch": 6.75, "learning_rate": 1.3066613322664533e-06, "loss": 0.0121, "step": 21743 }, { "epoch": 6.75, "learning_rate": 1.3062612522504502e-06, "loss": 0.0028, "step": 21744 }, { "epoch": 6.75, "learning_rate": 1.305861172234447e-06, "loss": 0.0032, "step": 21745 }, { "epoch": 6.75, "learning_rate": 1.3054610922184438e-06, "loss": 0.003, "step": 21746 }, { "epoch": 6.75, "learning_rate": 1.3050610122024407e-06, "loss": 0.0091, "step": 21747 }, { "epoch": 6.75, "learning_rate": 1.3046609321864374e-06, "loss": 0.0028, "step": 21748 }, { "epoch": 6.75, "learning_rate": 1.3042608521704343e-06, "loss": 0.0009, "step": 21749 }, { "epoch": 6.75, "learning_rate": 1.3038607721544312e-06, "loss": 0.0019, "step": 21750 }, { "epoch": 6.75, "learning_rate": 1.3034606921384277e-06, "loss": 0.0006, "step": 21751 }, { "epoch": 6.75, "learning_rate": 1.3030606121224246e-06, "loss": 0.0008, "step": 21752 }, { "epoch": 6.75, "learning_rate": 1.3026605321064214e-06, "loss": 0.0039, "step": 21753 }, { "epoch": 6.75, "learning_rate": 1.3022604520904181e-06, "loss": 0.0045, "step": 21754 }, { "epoch": 6.75, "learning_rate": 1.301860372074415e-06, "loss": 0.0094, "step": 21755 }, { "epoch": 6.75, "learning_rate": 1.301460292058412e-06, "loss": 0.0009, "step": 21756 }, { "epoch": 6.75, "learning_rate": 1.3010602120424084e-06, "loss": 0.0304, "step": 21757 }, { "epoch": 6.75, "learning_rate": 1.3006601320264053e-06, "loss": 0.0005, "step": 21758 }, { "epoch": 6.75, "learning_rate": 1.3002600520104022e-06, "loss": 0.0006, "step": 21759 }, { "epoch": 6.75, "learning_rate": 1.299859971994399e-06, "loss": 0.005, "step": 21760 }, { "epoch": 6.75, "learning_rate": 1.2994598919783958e-06, "loss": 0.0014, "step": 21761 }, { "epoch": 6.75, "learning_rate": 1.2990598119623927e-06, "loss": 0.0394, "step": 21762 }, { "epoch": 6.75, "learning_rate": 1.2986597319463896e-06, "loss": 0.0025, "step": 21763 }, { "epoch": 6.76, "learning_rate": 1.298259651930386e-06, "loss": 0.0011, "step": 21764 }, { "epoch": 6.76, "learning_rate": 1.297859571914383e-06, "loss": 0.0018, "step": 21765 }, { "epoch": 6.76, "learning_rate": 1.2974594918983799e-06, "loss": 0.0175, "step": 21766 }, { "epoch": 6.76, "learning_rate": 1.2970594118823765e-06, "loss": 0.0066, "step": 21767 }, { "epoch": 6.76, "learning_rate": 1.2966593318663734e-06, "loss": 0.0013, "step": 21768 }, { "epoch": 6.76, "learning_rate": 1.2962592518503703e-06, "loss": 0.0539, "step": 21769 }, { "epoch": 6.76, "learning_rate": 1.2958591718343668e-06, "loss": 0.0012, "step": 21770 }, { "epoch": 6.76, "learning_rate": 1.2954590918183637e-06, "loss": 0.0014, "step": 21771 }, { "epoch": 6.76, "learning_rate": 1.2950590118023606e-06, "loss": 0.0016, "step": 21772 }, { "epoch": 6.76, "learning_rate": 1.2946589317863573e-06, "loss": 0.0028, "step": 21773 }, { "epoch": 6.76, "learning_rate": 1.2942588517703542e-06, "loss": 0.0027, "step": 21774 }, { "epoch": 6.76, "learning_rate": 1.293858771754351e-06, "loss": 0.001, "step": 21775 }, { "epoch": 6.76, "learning_rate": 1.2934586917383476e-06, "loss": 0.0049, "step": 21776 }, { "epoch": 6.76, "learning_rate": 1.2930586117223445e-06, "loss": 0.0023, "step": 21777 }, { "epoch": 6.76, "learning_rate": 1.2926585317063414e-06, "loss": 0.0007, "step": 21778 }, { "epoch": 6.76, "learning_rate": 1.2922584516903383e-06, "loss": 0.0057, "step": 21779 }, { "epoch": 6.76, "learning_rate": 1.291858371674335e-06, "loss": 0.0016, "step": 21780 }, { "epoch": 6.76, "learning_rate": 1.2914582916583318e-06, "loss": 0.0052, "step": 21781 }, { "epoch": 6.76, "learning_rate": 1.2910582116423287e-06, "loss": 0.0032, "step": 21782 }, { "epoch": 6.76, "learning_rate": 1.2906581316263252e-06, "loss": 0.0784, "step": 21783 }, { "epoch": 6.76, "learning_rate": 1.2902580516103221e-06, "loss": 0.0083, "step": 21784 }, { "epoch": 6.76, "learning_rate": 1.289857971594319e-06, "loss": 0.0035, "step": 21785 }, { "epoch": 6.76, "learning_rate": 1.2894578915783157e-06, "loss": 0.0055, "step": 21786 }, { "epoch": 6.76, "learning_rate": 1.2890578115623126e-06, "loss": 0.0008, "step": 21787 }, { "epoch": 6.76, "learning_rate": 1.2886577315463095e-06, "loss": 0.0026, "step": 21788 }, { "epoch": 6.76, "learning_rate": 1.2882576515303062e-06, "loss": 0.0015, "step": 21789 }, { "epoch": 6.76, "learning_rate": 1.2878575715143029e-06, "loss": 0.0029, "step": 21790 }, { "epoch": 6.76, "learning_rate": 1.2874574914982998e-06, "loss": 0.0046, "step": 21791 }, { "epoch": 6.76, "learning_rate": 1.2870574114822964e-06, "loss": 0.001, "step": 21792 }, { "epoch": 6.76, "learning_rate": 1.2866573314662933e-06, "loss": 0.0012, "step": 21793 }, { "epoch": 6.76, "learning_rate": 1.2862572514502902e-06, "loss": 0.0275, "step": 21794 }, { "epoch": 6.76, "learning_rate": 1.285857171434287e-06, "loss": 0.0012, "step": 21795 }, { "epoch": 6.77, "learning_rate": 1.2854570914182838e-06, "loss": 0.0069, "step": 21796 }, { "epoch": 6.77, "learning_rate": 1.2850570114022805e-06, "loss": 0.0054, "step": 21797 }, { "epoch": 6.77, "learning_rate": 1.2846569313862772e-06, "loss": 0.0012, "step": 21798 }, { "epoch": 6.77, "learning_rate": 1.284256851370274e-06, "loss": 0.0216, "step": 21799 }, { "epoch": 6.77, "learning_rate": 1.283856771354271e-06, "loss": 0.0018, "step": 21800 }, { "epoch": 6.77, "learning_rate": 1.283456691338268e-06, "loss": 0.0006, "step": 21801 }, { "epoch": 6.77, "learning_rate": 1.2830566113222646e-06, "loss": 0.0012, "step": 21802 }, { "epoch": 6.77, "learning_rate": 1.2826565313062615e-06, "loss": 0.0008, "step": 21803 }, { "epoch": 6.77, "learning_rate": 1.2822564512902584e-06, "loss": 0.0006, "step": 21804 }, { "epoch": 6.77, "learning_rate": 1.2818563712742549e-06, "loss": 0.0101, "step": 21805 }, { "epoch": 6.77, "learning_rate": 1.2814562912582517e-06, "loss": 0.0016, "step": 21806 }, { "epoch": 6.77, "learning_rate": 1.2810562112422486e-06, "loss": 0.0012, "step": 21807 }, { "epoch": 6.77, "learning_rate": 1.2806561312262453e-06, "loss": 0.0088, "step": 21808 }, { "epoch": 6.77, "learning_rate": 1.2802560512102422e-06, "loss": 0.0004, "step": 21809 }, { "epoch": 6.77, "learning_rate": 1.2798559711942391e-06, "loss": 0.0023, "step": 21810 }, { "epoch": 6.77, "learning_rate": 1.2794558911782356e-06, "loss": 0.0015, "step": 21811 }, { "epoch": 6.77, "learning_rate": 1.2790558111622325e-06, "loss": 0.0019, "step": 21812 }, { "epoch": 6.77, "learning_rate": 1.2786557311462294e-06, "loss": 0.0025, "step": 21813 }, { "epoch": 6.77, "learning_rate": 1.278255651130226e-06, "loss": 0.0006, "step": 21814 }, { "epoch": 6.77, "learning_rate": 1.277855571114223e-06, "loss": 0.0011, "step": 21815 }, { "epoch": 6.77, "learning_rate": 1.2774554910982199e-06, "loss": 0.0013, "step": 21816 }, { "epoch": 6.77, "learning_rate": 1.2770554110822164e-06, "loss": 0.0012, "step": 21817 }, { "epoch": 6.77, "learning_rate": 1.2766553310662133e-06, "loss": 0.0012, "step": 21818 }, { "epoch": 6.77, "learning_rate": 1.2762552510502102e-06, "loss": 0.0012, "step": 21819 }, { "epoch": 6.77, "learning_rate": 1.2758551710342068e-06, "loss": 0.0008, "step": 21820 }, { "epoch": 6.77, "learning_rate": 1.2754550910182037e-06, "loss": 0.0023, "step": 21821 }, { "epoch": 6.77, "learning_rate": 1.2750550110022006e-06, "loss": 0.0036, "step": 21822 }, { "epoch": 6.77, "learning_rate": 1.2746549309861975e-06, "loss": 0.0147, "step": 21823 }, { "epoch": 6.77, "learning_rate": 1.274254850970194e-06, "loss": 0.0013, "step": 21824 }, { "epoch": 6.77, "learning_rate": 1.273854770954191e-06, "loss": 0.0035, "step": 21825 }, { "epoch": 6.77, "learning_rate": 1.2734546909381878e-06, "loss": 0.0012, "step": 21826 }, { "epoch": 6.77, "learning_rate": 1.2730546109221845e-06, "loss": 0.0013, "step": 21827 }, { "epoch": 6.77, "learning_rate": 1.2726545309061814e-06, "loss": 0.0038, "step": 21828 }, { "epoch": 6.78, "learning_rate": 1.2722544508901783e-06, "loss": 0.0022, "step": 21829 }, { "epoch": 6.78, "learning_rate": 1.2718543708741748e-06, "loss": 0.0018, "step": 21830 }, { "epoch": 6.78, "learning_rate": 1.2714542908581717e-06, "loss": 0.0031, "step": 21831 }, { "epoch": 6.78, "learning_rate": 1.2710542108421686e-06, "loss": 0.001, "step": 21832 }, { "epoch": 6.78, "learning_rate": 1.2706541308261652e-06, "loss": 0.0016, "step": 21833 }, { "epoch": 6.78, "learning_rate": 1.2702540508101621e-06, "loss": 0.0013, "step": 21834 }, { "epoch": 6.78, "learning_rate": 1.269853970794159e-06, "loss": 0.0025, "step": 21835 }, { "epoch": 6.78, "learning_rate": 1.2694538907781555e-06, "loss": 0.0606, "step": 21836 }, { "epoch": 6.78, "learning_rate": 1.2690538107621524e-06, "loss": 0.0007, "step": 21837 }, { "epoch": 6.78, "learning_rate": 1.2686537307461493e-06, "loss": 0.0032, "step": 21838 }, { "epoch": 6.78, "learning_rate": 1.268253650730146e-06, "loss": 0.0023, "step": 21839 }, { "epoch": 6.78, "learning_rate": 1.267853570714143e-06, "loss": 0.0032, "step": 21840 }, { "epoch": 6.78, "learning_rate": 1.2674534906981398e-06, "loss": 0.0014, "step": 21841 }, { "epoch": 6.78, "learning_rate": 1.2670534106821367e-06, "loss": 0.0015, "step": 21842 }, { "epoch": 6.78, "learning_rate": 1.2666533306661334e-06, "loss": 0.0014, "step": 21843 }, { "epoch": 6.78, "learning_rate": 1.26625325065013e-06, "loss": 0.0022, "step": 21844 }, { "epoch": 6.78, "learning_rate": 1.265853170634127e-06, "loss": 0.0006, "step": 21845 }, { "epoch": 6.78, "learning_rate": 1.2654530906181236e-06, "loss": 0.0039, "step": 21846 }, { "epoch": 6.78, "learning_rate": 1.2650530106021205e-06, "loss": 0.001, "step": 21847 }, { "epoch": 6.78, "learning_rate": 1.2646529305861174e-06, "loss": 0.0015, "step": 21848 }, { "epoch": 6.78, "learning_rate": 1.2642528505701141e-06, "loss": 0.0015, "step": 21849 }, { "epoch": 6.78, "learning_rate": 1.263852770554111e-06, "loss": 0.0012, "step": 21850 }, { "epoch": 6.78, "learning_rate": 1.2634526905381077e-06, "loss": 0.0043, "step": 21851 }, { "epoch": 6.78, "learning_rate": 1.2630526105221044e-06, "loss": 0.0009, "step": 21852 }, { "epoch": 6.78, "learning_rate": 1.2626525305061013e-06, "loss": 0.0107, "step": 21853 }, { "epoch": 6.78, "learning_rate": 1.2622524504900982e-06, "loss": 0.0037, "step": 21854 }, { "epoch": 6.78, "learning_rate": 1.2618523704740949e-06, "loss": 0.0018, "step": 21855 }, { "epoch": 6.78, "learning_rate": 1.2614522904580918e-06, "loss": 0.0008, "step": 21856 }, { "epoch": 6.78, "learning_rate": 1.2610522104420887e-06, "loss": 0.0073, "step": 21857 }, { "epoch": 6.78, "learning_rate": 1.2606521304260852e-06, "loss": 0.0139, "step": 21858 }, { "epoch": 6.78, "learning_rate": 1.260252050410082e-06, "loss": 0.0159, "step": 21859 }, { "epoch": 6.78, "learning_rate": 1.259851970394079e-06, "loss": 0.0035, "step": 21860 }, { "epoch": 6.79, "learning_rate": 1.2594518903780756e-06, "loss": 0.0108, "step": 21861 }, { "epoch": 6.79, "learning_rate": 1.2590518103620725e-06, "loss": 0.0013, "step": 21862 }, { "epoch": 6.79, "learning_rate": 1.2586517303460694e-06, "loss": 0.0127, "step": 21863 }, { "epoch": 6.79, "learning_rate": 1.2582516503300663e-06, "loss": 0.0057, "step": 21864 }, { "epoch": 6.79, "learning_rate": 1.2578515703140628e-06, "loss": 0.0065, "step": 21865 }, { "epoch": 6.79, "learning_rate": 1.2574514902980597e-06, "loss": 0.0012, "step": 21866 }, { "epoch": 6.79, "learning_rate": 1.2570514102820566e-06, "loss": 0.001, "step": 21867 }, { "epoch": 6.79, "learning_rate": 1.2566513302660533e-06, "loss": 0.0242, "step": 21868 }, { "epoch": 6.79, "learning_rate": 1.2562512502500502e-06, "loss": 0.0329, "step": 21869 }, { "epoch": 6.79, "learning_rate": 1.255851170234047e-06, "loss": 0.0012, "step": 21870 }, { "epoch": 6.79, "learning_rate": 1.2554510902180436e-06, "loss": 0.0014, "step": 21871 }, { "epoch": 6.79, "learning_rate": 1.2550510102020405e-06, "loss": 0.0009, "step": 21872 }, { "epoch": 6.79, "learning_rate": 1.2546509301860374e-06, "loss": 0.0011, "step": 21873 }, { "epoch": 6.79, "learning_rate": 1.254250850170034e-06, "loss": 0.0041, "step": 21874 }, { "epoch": 6.79, "learning_rate": 1.253850770154031e-06, "loss": 0.0016, "step": 21875 }, { "epoch": 6.79, "learning_rate": 1.2534506901380278e-06, "loss": 0.0078, "step": 21876 }, { "epoch": 6.79, "learning_rate": 1.2530506101220243e-06, "loss": 0.002, "step": 21877 }, { "epoch": 6.79, "learning_rate": 1.2526505301060212e-06, "loss": 0.0057, "step": 21878 }, { "epoch": 6.79, "learning_rate": 1.2522504500900181e-06, "loss": 0.0023, "step": 21879 }, { "epoch": 6.79, "learning_rate": 1.2518503700740148e-06, "loss": 0.0012, "step": 21880 }, { "epoch": 6.79, "learning_rate": 1.2514502900580117e-06, "loss": 0.001, "step": 21881 }, { "epoch": 6.79, "learning_rate": 1.2510502100420086e-06, "loss": 0.0024, "step": 21882 }, { "epoch": 6.79, "learning_rate": 1.2506501300260055e-06, "loss": 0.0007, "step": 21883 }, { "epoch": 6.79, "learning_rate": 1.250250050010002e-06, "loss": 0.0017, "step": 21884 }, { "epoch": 6.79, "learning_rate": 1.2498499699939989e-06, "loss": 0.002, "step": 21885 }, { "epoch": 6.79, "learning_rate": 1.2494498899779958e-06, "loss": 0.0427, "step": 21886 }, { "epoch": 6.79, "learning_rate": 1.2490498099619924e-06, "loss": 0.0027, "step": 21887 }, { "epoch": 6.79, "learning_rate": 1.2486497299459893e-06, "loss": 0.0015, "step": 21888 }, { "epoch": 6.79, "learning_rate": 1.248249649929986e-06, "loss": 0.0128, "step": 21889 }, { "epoch": 6.79, "learning_rate": 1.247849569913983e-06, "loss": 0.0019, "step": 21890 }, { "epoch": 6.79, "learning_rate": 1.2474494898979796e-06, "loss": 0.0019, "step": 21891 }, { "epoch": 6.79, "learning_rate": 1.2470494098819765e-06, "loss": 0.0309, "step": 21892 }, { "epoch": 6.8, "learning_rate": 1.2466493298659734e-06, "loss": 0.0016, "step": 21893 }, { "epoch": 6.8, "learning_rate": 1.24624924984997e-06, "loss": 0.0187, "step": 21894 }, { "epoch": 6.8, "learning_rate": 1.2458491698339668e-06, "loss": 0.0035, "step": 21895 }, { "epoch": 6.8, "learning_rate": 1.2454490898179637e-06, "loss": 0.001, "step": 21896 }, { "epoch": 6.8, "learning_rate": 1.2450490098019606e-06, "loss": 0.0006, "step": 21897 }, { "epoch": 6.8, "learning_rate": 1.2446489297859573e-06, "loss": 0.0077, "step": 21898 }, { "epoch": 6.8, "learning_rate": 1.2442488497699542e-06, "loss": 0.0029, "step": 21899 }, { "epoch": 6.8, "learning_rate": 1.2438487697539508e-06, "loss": 0.0021, "step": 21900 }, { "epoch": 6.8, "learning_rate": 1.2434486897379477e-06, "loss": 0.0008, "step": 21901 }, { "epoch": 6.8, "learning_rate": 1.2430486097219444e-06, "loss": 0.0018, "step": 21902 }, { "epoch": 6.8, "learning_rate": 1.2426485297059413e-06, "loss": 0.0015, "step": 21903 }, { "epoch": 6.8, "learning_rate": 1.2422484496899382e-06, "loss": 0.0025, "step": 21904 }, { "epoch": 6.8, "learning_rate": 1.241848369673935e-06, "loss": 0.0183, "step": 21905 }, { "epoch": 6.8, "learning_rate": 1.2414482896579316e-06, "loss": 0.0024, "step": 21906 }, { "epoch": 6.8, "learning_rate": 1.2410482096419285e-06, "loss": 0.0015, "step": 21907 }, { "epoch": 6.8, "learning_rate": 1.2406481296259252e-06, "loss": 0.0016, "step": 21908 }, { "epoch": 6.8, "learning_rate": 1.240248049609922e-06, "loss": 0.0006, "step": 21909 }, { "epoch": 6.8, "learning_rate": 1.239847969593919e-06, "loss": 0.0007, "step": 21910 }, { "epoch": 6.8, "learning_rate": 1.2394478895779157e-06, "loss": 0.0014, "step": 21911 }, { "epoch": 6.8, "learning_rate": 1.2390478095619126e-06, "loss": 0.001, "step": 21912 }, { "epoch": 6.8, "learning_rate": 1.2386477295459093e-06, "loss": 0.0016, "step": 21913 }, { "epoch": 6.8, "learning_rate": 1.238247649529906e-06, "loss": 0.0027, "step": 21914 }, { "epoch": 6.8, "learning_rate": 1.2378475695139028e-06, "loss": 0.0276, "step": 21915 }, { "epoch": 6.8, "learning_rate": 1.2374474894978997e-06, "loss": 0.0021, "step": 21916 }, { "epoch": 6.8, "learning_rate": 1.2370474094818964e-06, "loss": 0.0033, "step": 21917 }, { "epoch": 6.8, "learning_rate": 1.2366473294658933e-06, "loss": 0.0038, "step": 21918 }, { "epoch": 6.8, "learning_rate": 1.23624724944989e-06, "loss": 0.0017, "step": 21919 }, { "epoch": 6.8, "learning_rate": 1.235847169433887e-06, "loss": 0.0042, "step": 21920 }, { "epoch": 6.8, "learning_rate": 1.2354470894178836e-06, "loss": 0.0029, "step": 21921 }, { "epoch": 6.8, "learning_rate": 1.2350470094018805e-06, "loss": 0.0014, "step": 21922 }, { "epoch": 6.8, "learning_rate": 1.2346469293858774e-06, "loss": 0.0019, "step": 21923 }, { "epoch": 6.8, "learning_rate": 1.234246849369874e-06, "loss": 0.0026, "step": 21924 }, { "epoch": 6.81, "learning_rate": 1.2338467693538708e-06, "loss": 0.0274, "step": 21925 }, { "epoch": 6.81, "learning_rate": 1.2334466893378677e-06, "loss": 0.0017, "step": 21926 }, { "epoch": 6.81, "learning_rate": 1.2330466093218646e-06, "loss": 0.0022, "step": 21927 }, { "epoch": 6.81, "learning_rate": 1.2326465293058612e-06, "loss": 0.0515, "step": 21928 }, { "epoch": 6.81, "learning_rate": 1.2322464492898581e-06, "loss": 0.0006, "step": 21929 }, { "epoch": 6.81, "learning_rate": 1.2318463692738548e-06, "loss": 0.0086, "step": 21930 }, { "epoch": 6.81, "learning_rate": 1.2314462892578517e-06, "loss": 0.0014, "step": 21931 }, { "epoch": 6.81, "learning_rate": 1.2310462092418484e-06, "loss": 0.0016, "step": 21932 }, { "epoch": 6.81, "learning_rate": 1.2306461292258453e-06, "loss": 0.0009, "step": 21933 }, { "epoch": 6.81, "learning_rate": 1.2302460492098422e-06, "loss": 0.0007, "step": 21934 }, { "epoch": 6.81, "learning_rate": 1.2298459691938389e-06, "loss": 0.026, "step": 21935 }, { "epoch": 6.81, "learning_rate": 1.2294458891778356e-06, "loss": 0.0021, "step": 21936 }, { "epoch": 6.81, "learning_rate": 1.2290458091618325e-06, "loss": 0.0016, "step": 21937 }, { "epoch": 6.81, "learning_rate": 1.2286457291458292e-06, "loss": 0.001, "step": 21938 }, { "epoch": 6.81, "learning_rate": 1.228245649129826e-06, "loss": 0.0008, "step": 21939 }, { "epoch": 6.81, "learning_rate": 1.227845569113823e-06, "loss": 0.002, "step": 21940 }, { "epoch": 6.81, "learning_rate": 1.2274454890978196e-06, "loss": 0.0012, "step": 21941 }, { "epoch": 6.81, "learning_rate": 1.2270454090818165e-06, "loss": 0.0013, "step": 21942 }, { "epoch": 6.81, "learning_rate": 1.2266453290658132e-06, "loss": 0.0019, "step": 21943 }, { "epoch": 6.81, "learning_rate": 1.22624524904981e-06, "loss": 0.001, "step": 21944 }, { "epoch": 6.81, "learning_rate": 1.2258451690338068e-06, "loss": 0.0012, "step": 21945 }, { "epoch": 6.81, "learning_rate": 1.2254450890178037e-06, "loss": 0.0015, "step": 21946 }, { "epoch": 6.81, "learning_rate": 1.2250450090018004e-06, "loss": 0.0009, "step": 21947 }, { "epoch": 6.81, "learning_rate": 1.2246449289857973e-06, "loss": 0.0018, "step": 21948 }, { "epoch": 6.81, "learning_rate": 1.224244848969794e-06, "loss": 0.0015, "step": 21949 }, { "epoch": 6.81, "learning_rate": 1.2238447689537909e-06, "loss": 0.0011, "step": 21950 }, { "epoch": 6.81, "learning_rate": 1.2234446889377878e-06, "loss": 0.0029, "step": 21951 }, { "epoch": 6.81, "learning_rate": 1.2230446089217845e-06, "loss": 0.0032, "step": 21952 }, { "epoch": 6.81, "learning_rate": 1.2226445289057814e-06, "loss": 0.0014, "step": 21953 }, { "epoch": 6.81, "learning_rate": 1.222244448889778e-06, "loss": 0.0015, "step": 21954 }, { "epoch": 6.81, "learning_rate": 1.2218443688737747e-06, "loss": 0.0021, "step": 21955 }, { "epoch": 6.81, "learning_rate": 1.2214442888577716e-06, "loss": 0.0027, "step": 21956 }, { "epoch": 6.81, "learning_rate": 1.2210442088417685e-06, "loss": 0.0015, "step": 21957 }, { "epoch": 6.82, "learning_rate": 1.2206441288257652e-06, "loss": 0.0013, "step": 21958 }, { "epoch": 6.82, "learning_rate": 1.2202440488097621e-06, "loss": 0.0012, "step": 21959 }, { "epoch": 6.82, "learning_rate": 1.2198439687937588e-06, "loss": 0.0009, "step": 21960 }, { "epoch": 6.82, "learning_rate": 1.2194438887777557e-06, "loss": 0.0405, "step": 21961 }, { "epoch": 6.82, "learning_rate": 1.2190438087617524e-06, "loss": 0.0191, "step": 21962 }, { "epoch": 6.82, "learning_rate": 1.2186437287457493e-06, "loss": 0.0051, "step": 21963 }, { "epoch": 6.82, "learning_rate": 1.2182436487297462e-06, "loss": 0.0013, "step": 21964 }, { "epoch": 6.82, "learning_rate": 1.2178435687137429e-06, "loss": 0.0023, "step": 21965 }, { "epoch": 6.82, "learning_rate": 1.2174434886977396e-06, "loss": 0.0006, "step": 21966 }, { "epoch": 6.82, "learning_rate": 1.2170434086817365e-06, "loss": 0.0026, "step": 21967 }, { "epoch": 6.82, "learning_rate": 1.2166433286657331e-06, "loss": 0.0121, "step": 21968 }, { "epoch": 6.82, "learning_rate": 1.21624324864973e-06, "loss": 0.0009, "step": 21969 }, { "epoch": 6.82, "learning_rate": 1.215843168633727e-06, "loss": 0.0011, "step": 21970 }, { "epoch": 6.82, "learning_rate": 1.2154430886177236e-06, "loss": 0.0032, "step": 21971 }, { "epoch": 6.82, "learning_rate": 1.2150430086017205e-06, "loss": 0.0077, "step": 21972 }, { "epoch": 6.82, "learning_rate": 1.2146429285857172e-06, "loss": 0.0013, "step": 21973 }, { "epoch": 6.82, "learning_rate": 1.214242848569714e-06, "loss": 0.0027, "step": 21974 }, { "epoch": 6.82, "learning_rate": 1.2138427685537108e-06, "loss": 0.0016, "step": 21975 }, { "epoch": 6.82, "learning_rate": 1.2134426885377077e-06, "loss": 0.0013, "step": 21976 }, { "epoch": 6.82, "learning_rate": 1.2130426085217044e-06, "loss": 0.0043, "step": 21977 }, { "epoch": 6.82, "learning_rate": 1.2126425285057013e-06, "loss": 0.0012, "step": 21978 }, { "epoch": 6.82, "learning_rate": 1.212242448489698e-06, "loss": 0.006, "step": 21979 }, { "epoch": 6.82, "learning_rate": 1.2118423684736949e-06, "loss": 0.001, "step": 21980 }, { "epoch": 6.82, "learning_rate": 1.2114422884576918e-06, "loss": 0.0015, "step": 21981 }, { "epoch": 6.82, "learning_rate": 1.2110422084416884e-06, "loss": 0.0065, "step": 21982 }, { "epoch": 6.82, "learning_rate": 1.2106421284256853e-06, "loss": 0.0012, "step": 21983 }, { "epoch": 6.82, "learning_rate": 1.210242048409682e-06, "loss": 0.0043, "step": 21984 }, { "epoch": 6.82, "learning_rate": 1.2098419683936787e-06, "loss": 0.0009, "step": 21985 }, { "epoch": 6.82, "learning_rate": 1.2094418883776756e-06, "loss": 0.0025, "step": 21986 }, { "epoch": 6.82, "learning_rate": 1.2090418083616725e-06, "loss": 0.0013, "step": 21987 }, { "epoch": 6.82, "learning_rate": 1.2086417283456692e-06, "loss": 0.001, "step": 21988 }, { "epoch": 6.82, "learning_rate": 1.208241648329666e-06, "loss": 0.0009, "step": 21989 }, { "epoch": 6.83, "learning_rate": 1.2078415683136628e-06, "loss": 0.0097, "step": 21990 }, { "epoch": 6.83, "learning_rate": 1.2074414882976597e-06, "loss": 0.0011, "step": 21991 }, { "epoch": 6.83, "learning_rate": 1.2070414082816564e-06, "loss": 0.0013, "step": 21992 }, { "epoch": 6.83, "learning_rate": 1.2066413282656533e-06, "loss": 0.0011, "step": 21993 }, { "epoch": 6.83, "learning_rate": 1.2062412482496502e-06, "loss": 0.0007, "step": 21994 }, { "epoch": 6.83, "learning_rate": 1.2058411682336468e-06, "loss": 0.0031, "step": 21995 }, { "epoch": 6.83, "learning_rate": 1.2054410882176435e-06, "loss": 0.0024, "step": 21996 }, { "epoch": 6.83, "learning_rate": 1.2050410082016404e-06, "loss": 0.0014, "step": 21997 }, { "epoch": 6.83, "learning_rate": 1.2046409281856371e-06, "loss": 0.0016, "step": 21998 }, { "epoch": 6.83, "learning_rate": 1.204240848169634e-06, "loss": 0.0012, "step": 21999 }, { "epoch": 6.83, "learning_rate": 1.203840768153631e-06, "loss": 0.0083, "step": 22000 }, { "epoch": 6.83, "learning_rate": 1.2034406881376276e-06, "loss": 0.0092, "step": 22001 }, { "epoch": 6.83, "learning_rate": 1.2030406081216245e-06, "loss": 0.0013, "step": 22002 }, { "epoch": 6.83, "learning_rate": 1.2026405281056212e-06, "loss": 0.0019, "step": 22003 }, { "epoch": 6.83, "learning_rate": 1.202240448089618e-06, "loss": 0.0349, "step": 22004 }, { "epoch": 6.83, "learning_rate": 1.201840368073615e-06, "loss": 0.0017, "step": 22005 }, { "epoch": 6.83, "learning_rate": 1.2014402880576117e-06, "loss": 0.0023, "step": 22006 }, { "epoch": 6.83, "learning_rate": 1.2010402080416083e-06, "loss": 0.0193, "step": 22007 }, { "epoch": 6.83, "learning_rate": 1.2006401280256052e-06, "loss": 0.0011, "step": 22008 }, { "epoch": 6.83, "learning_rate": 1.200240048009602e-06, "loss": 0.0045, "step": 22009 }, { "epoch": 6.83, "learning_rate": 1.1998399679935988e-06, "loss": 0.0013, "step": 22010 }, { "epoch": 6.83, "learning_rate": 1.1994398879775957e-06, "loss": 0.0061, "step": 22011 }, { "epoch": 6.83, "learning_rate": 1.1990398079615924e-06, "loss": 0.0009, "step": 22012 }, { "epoch": 6.83, "learning_rate": 1.1986397279455893e-06, "loss": 0.0006, "step": 22013 }, { "epoch": 6.83, "learning_rate": 1.198239647929586e-06, "loss": 0.0134, "step": 22014 }, { "epoch": 6.83, "learning_rate": 1.1978395679135827e-06, "loss": 0.0008, "step": 22015 }, { "epoch": 6.83, "learning_rate": 1.1974394878975796e-06, "loss": 0.0006, "step": 22016 }, { "epoch": 6.83, "learning_rate": 1.1970394078815765e-06, "loss": 0.0013, "step": 22017 }, { "epoch": 6.83, "learning_rate": 1.1966393278655732e-06, "loss": 0.0312, "step": 22018 }, { "epoch": 6.83, "learning_rate": 1.19623924784957e-06, "loss": 0.0034, "step": 22019 }, { "epoch": 6.83, "learning_rate": 1.1958391678335668e-06, "loss": 0.0369, "step": 22020 }, { "epoch": 6.83, "learning_rate": 1.1954390878175634e-06, "loss": 0.0024, "step": 22021 }, { "epoch": 6.84, "learning_rate": 1.1950390078015603e-06, "loss": 0.0013, "step": 22022 }, { "epoch": 6.84, "learning_rate": 1.1946389277855572e-06, "loss": 0.0015, "step": 22023 }, { "epoch": 6.84, "learning_rate": 1.1942388477695541e-06, "loss": 0.0042, "step": 22024 }, { "epoch": 6.84, "learning_rate": 1.1938387677535508e-06, "loss": 0.0015, "step": 22025 }, { "epoch": 6.84, "learning_rate": 1.1934386877375475e-06, "loss": 0.025, "step": 22026 }, { "epoch": 6.84, "learning_rate": 1.1930386077215444e-06, "loss": 0.0011, "step": 22027 }, { "epoch": 6.84, "learning_rate": 1.1926385277055413e-06, "loss": 0.001, "step": 22028 }, { "epoch": 6.84, "learning_rate": 1.192238447689538e-06, "loss": 0.0012, "step": 22029 }, { "epoch": 6.84, "learning_rate": 1.1918383676735349e-06, "loss": 0.0025, "step": 22030 }, { "epoch": 6.84, "learning_rate": 1.1914382876575316e-06, "loss": 0.0011, "step": 22031 }, { "epoch": 6.84, "learning_rate": 1.1910382076415285e-06, "loss": 0.0035, "step": 22032 }, { "epoch": 6.84, "learning_rate": 1.1906381276255252e-06, "loss": 0.0074, "step": 22033 }, { "epoch": 6.84, "learning_rate": 1.190238047609522e-06, "loss": 0.0053, "step": 22034 }, { "epoch": 6.84, "learning_rate": 1.189837967593519e-06, "loss": 0.0014, "step": 22035 }, { "epoch": 6.84, "learning_rate": 1.1894378875775156e-06, "loss": 0.002, "step": 22036 }, { "epoch": 6.84, "learning_rate": 1.1890378075615123e-06, "loss": 0.0025, "step": 22037 }, { "epoch": 6.84, "learning_rate": 1.1886377275455092e-06, "loss": 0.006, "step": 22038 }, { "epoch": 6.84, "learning_rate": 1.188237647529506e-06, "loss": 0.0014, "step": 22039 }, { "epoch": 6.84, "learning_rate": 1.1878375675135028e-06, "loss": 0.0022, "step": 22040 }, { "epoch": 6.84, "learning_rate": 1.1874374874974997e-06, "loss": 0.0014, "step": 22041 }, { "epoch": 6.84, "learning_rate": 1.1870374074814964e-06, "loss": 0.0125, "step": 22042 }, { "epoch": 6.84, "learning_rate": 1.1866373274654933e-06, "loss": 0.0014, "step": 22043 }, { "epoch": 6.84, "learning_rate": 1.18623724744949e-06, "loss": 0.0013, "step": 22044 }, { "epoch": 6.84, "learning_rate": 1.1858371674334867e-06, "loss": 0.002, "step": 22045 }, { "epoch": 6.84, "learning_rate": 1.1854370874174836e-06, "loss": 0.0062, "step": 22046 }, { "epoch": 6.84, "learning_rate": 1.1850370074014805e-06, "loss": 0.0088, "step": 22047 }, { "epoch": 6.84, "learning_rate": 1.1846369273854771e-06, "loss": 0.0017, "step": 22048 }, { "epoch": 6.84, "learning_rate": 1.184236847369474e-06, "loss": 0.0041, "step": 22049 }, { "epoch": 6.84, "learning_rate": 1.1838367673534707e-06, "loss": 0.008, "step": 22050 }, { "epoch": 6.84, "learning_rate": 1.1834366873374676e-06, "loss": 0.0013, "step": 22051 }, { "epoch": 6.84, "learning_rate": 1.1830366073214643e-06, "loss": 0.0017, "step": 22052 }, { "epoch": 6.84, "learning_rate": 1.1826365273054612e-06, "loss": 0.0017, "step": 22053 }, { "epoch": 6.85, "learning_rate": 1.1822364472894581e-06, "loss": 0.0027, "step": 22054 }, { "epoch": 6.85, "learning_rate": 1.1818363672734548e-06, "loss": 0.0365, "step": 22055 }, { "epoch": 6.85, "learning_rate": 1.1814362872574515e-06, "loss": 0.0014, "step": 22056 }, { "epoch": 6.85, "learning_rate": 1.1810362072414484e-06, "loss": 0.0023, "step": 22057 }, { "epoch": 6.85, "learning_rate": 1.1806361272254453e-06, "loss": 0.0008, "step": 22058 }, { "epoch": 6.85, "learning_rate": 1.180236047209442e-06, "loss": 0.0017, "step": 22059 }, { "epoch": 6.85, "learning_rate": 1.1798359671934389e-06, "loss": 0.0012, "step": 22060 }, { "epoch": 6.85, "learning_rate": 1.1794358871774355e-06, "loss": 0.0034, "step": 22061 }, { "epoch": 6.85, "learning_rate": 1.1790358071614322e-06, "loss": 0.0144, "step": 22062 }, { "epoch": 6.85, "learning_rate": 1.1786357271454291e-06, "loss": 0.0009, "step": 22063 }, { "epoch": 6.85, "learning_rate": 1.178235647129426e-06, "loss": 0.0009, "step": 22064 }, { "epoch": 6.85, "learning_rate": 1.177835567113423e-06, "loss": 0.0014, "step": 22065 }, { "epoch": 6.85, "learning_rate": 1.1774354870974196e-06, "loss": 0.0092, "step": 22066 }, { "epoch": 6.85, "learning_rate": 1.1770354070814163e-06, "loss": 0.0012, "step": 22067 }, { "epoch": 6.85, "learning_rate": 1.1766353270654132e-06, "loss": 0.0033, "step": 22068 }, { "epoch": 6.85, "learning_rate": 1.1762352470494099e-06, "loss": 0.0012, "step": 22069 }, { "epoch": 6.85, "learning_rate": 1.1758351670334068e-06, "loss": 0.001, "step": 22070 }, { "epoch": 6.85, "learning_rate": 1.1754350870174037e-06, "loss": 0.0043, "step": 22071 }, { "epoch": 6.85, "learning_rate": 1.1750350070014004e-06, "loss": 0.0026, "step": 22072 }, { "epoch": 6.85, "learning_rate": 1.1746349269853973e-06, "loss": 0.0335, "step": 22073 }, { "epoch": 6.85, "learning_rate": 1.174234846969394e-06, "loss": 0.0011, "step": 22074 }, { "epoch": 6.85, "learning_rate": 1.1738347669533906e-06, "loss": 0.0022, "step": 22075 }, { "epoch": 6.85, "learning_rate": 1.1734346869373875e-06, "loss": 0.0008, "step": 22076 }, { "epoch": 6.85, "learning_rate": 1.1730346069213844e-06, "loss": 0.0025, "step": 22077 }, { "epoch": 6.85, "learning_rate": 1.1726345269053811e-06, "loss": 0.0023, "step": 22078 }, { "epoch": 6.85, "learning_rate": 1.172234446889378e-06, "loss": 0.0013, "step": 22079 }, { "epoch": 6.85, "learning_rate": 1.1718343668733747e-06, "loss": 0.0545, "step": 22080 }, { "epoch": 6.85, "learning_rate": 1.1714342868573716e-06, "loss": 0.0011, "step": 22081 }, { "epoch": 6.85, "learning_rate": 1.1710342068413685e-06, "loss": 0.0011, "step": 22082 }, { "epoch": 6.85, "learning_rate": 1.1706341268253652e-06, "loss": 0.0029, "step": 22083 }, { "epoch": 6.85, "learning_rate": 1.170234046809362e-06, "loss": 0.0028, "step": 22084 }, { "epoch": 6.85, "learning_rate": 1.1698339667933588e-06, "loss": 0.003, "step": 22085 }, { "epoch": 6.86, "learning_rate": 1.1694338867773555e-06, "loss": 0.0027, "step": 22086 }, { "epoch": 6.86, "learning_rate": 1.1690338067613524e-06, "loss": 0.0004, "step": 22087 }, { "epoch": 6.86, "learning_rate": 1.1686337267453493e-06, "loss": 0.0012, "step": 22088 }, { "epoch": 6.86, "learning_rate": 1.168233646729346e-06, "loss": 0.0024, "step": 22089 }, { "epoch": 6.86, "learning_rate": 1.1678335667133428e-06, "loss": 0.0101, "step": 22090 }, { "epoch": 6.86, "learning_rate": 1.1674334866973395e-06, "loss": 0.0019, "step": 22091 }, { "epoch": 6.86, "learning_rate": 1.1670334066813362e-06, "loss": 0.0015, "step": 22092 }, { "epoch": 6.86, "learning_rate": 1.1666333266653331e-06, "loss": 0.0008, "step": 22093 }, { "epoch": 6.86, "learning_rate": 1.16623324664933e-06, "loss": 0.0011, "step": 22094 }, { "epoch": 6.86, "learning_rate": 1.165833166633327e-06, "loss": 0.0009, "step": 22095 }, { "epoch": 6.86, "learning_rate": 1.1654330866173236e-06, "loss": 0.0014, "step": 22096 }, { "epoch": 6.86, "learning_rate": 1.1650330066013203e-06, "loss": 0.001, "step": 22097 }, { "epoch": 6.86, "learning_rate": 1.1646329265853172e-06, "loss": 0.0025, "step": 22098 }, { "epoch": 6.86, "learning_rate": 1.1642328465693139e-06, "loss": 0.0114, "step": 22099 }, { "epoch": 6.86, "learning_rate": 1.1638327665533108e-06, "loss": 0.0011, "step": 22100 }, { "epoch": 6.86, "learning_rate": 1.1634326865373077e-06, "loss": 0.0011, "step": 22101 }, { "epoch": 6.86, "learning_rate": 1.1630326065213043e-06, "loss": 0.0105, "step": 22102 }, { "epoch": 6.86, "learning_rate": 1.162632526505301e-06, "loss": 0.001, "step": 22103 }, { "epoch": 6.86, "learning_rate": 1.162232446489298e-06, "loss": 0.0021, "step": 22104 }, { "epoch": 6.86, "learning_rate": 1.1618323664732948e-06, "loss": 0.0196, "step": 22105 }, { "epoch": 6.86, "learning_rate": 1.1614322864572915e-06, "loss": 0.0006, "step": 22106 }, { "epoch": 6.86, "learning_rate": 1.1610322064412884e-06, "loss": 0.003, "step": 22107 }, { "epoch": 6.86, "learning_rate": 1.160632126425285e-06, "loss": 0.0019, "step": 22108 }, { "epoch": 6.86, "learning_rate": 1.160232046409282e-06, "loss": 0.0017, "step": 22109 }, { "epoch": 6.86, "learning_rate": 1.1598319663932787e-06, "loss": 0.0008, "step": 22110 }, { "epoch": 6.86, "learning_rate": 1.1594318863772756e-06, "loss": 0.0009, "step": 22111 }, { "epoch": 6.86, "learning_rate": 1.1590318063612725e-06, "loss": 0.0018, "step": 22112 }, { "epoch": 6.86, "learning_rate": 1.1586317263452692e-06, "loss": 0.0036, "step": 22113 }, { "epoch": 6.86, "learning_rate": 1.1582316463292659e-06, "loss": 0.0034, "step": 22114 }, { "epoch": 6.86, "learning_rate": 1.1578315663132627e-06, "loss": 0.0033, "step": 22115 }, { "epoch": 6.86, "learning_rate": 1.1574314862972594e-06, "loss": 0.0041, "step": 22116 }, { "epoch": 6.86, "learning_rate": 1.1570314062812563e-06, "loss": 0.03, "step": 22117 }, { "epoch": 6.86, "learning_rate": 1.1566313262652532e-06, "loss": 0.0404, "step": 22118 }, { "epoch": 6.87, "learning_rate": 1.15623124624925e-06, "loss": 0.0015, "step": 22119 }, { "epoch": 6.87, "learning_rate": 1.1558311662332468e-06, "loss": 0.0017, "step": 22120 }, { "epoch": 6.87, "learning_rate": 1.1554310862172435e-06, "loss": 0.0017, "step": 22121 }, { "epoch": 6.87, "learning_rate": 1.1550310062012402e-06, "loss": 0.0007, "step": 22122 }, { "epoch": 6.87, "learning_rate": 1.154630926185237e-06, "loss": 0.0021, "step": 22123 }, { "epoch": 6.87, "learning_rate": 1.154230846169234e-06, "loss": 0.0013, "step": 22124 }, { "epoch": 6.87, "learning_rate": 1.1538307661532309e-06, "loss": 0.0009, "step": 22125 }, { "epoch": 6.87, "learning_rate": 1.1534306861372276e-06, "loss": 0.0016, "step": 22126 }, { "epoch": 6.87, "learning_rate": 1.1530306061212243e-06, "loss": 0.0011, "step": 22127 }, { "epoch": 6.87, "learning_rate": 1.1526305261052212e-06, "loss": 0.0012, "step": 22128 }, { "epoch": 6.87, "learning_rate": 1.1522304460892178e-06, "loss": 0.0019, "step": 22129 }, { "epoch": 6.87, "learning_rate": 1.1518303660732147e-06, "loss": 0.002, "step": 22130 }, { "epoch": 6.87, "learning_rate": 1.1514302860572116e-06, "loss": 0.0009, "step": 22131 }, { "epoch": 6.87, "learning_rate": 1.1510302060412083e-06, "loss": 0.0012, "step": 22132 }, { "epoch": 6.87, "learning_rate": 1.150630126025205e-06, "loss": 0.0183, "step": 22133 }, { "epoch": 6.87, "learning_rate": 1.150230046009202e-06, "loss": 0.0023, "step": 22134 }, { "epoch": 6.87, "learning_rate": 1.1498299659931988e-06, "loss": 0.0028, "step": 22135 }, { "epoch": 6.87, "learning_rate": 1.1494298859771957e-06, "loss": 0.0012, "step": 22136 }, { "epoch": 6.87, "learning_rate": 1.1490298059611924e-06, "loss": 0.0013, "step": 22137 }, { "epoch": 6.87, "learning_rate": 1.148629725945189e-06, "loss": 0.0019, "step": 22138 }, { "epoch": 6.87, "learning_rate": 1.148229645929186e-06, "loss": 0.0012, "step": 22139 }, { "epoch": 6.87, "learning_rate": 1.1478295659131827e-06, "loss": 0.002, "step": 22140 }, { "epoch": 6.87, "learning_rate": 1.1474294858971796e-06, "loss": 0.0021, "step": 22141 }, { "epoch": 6.87, "learning_rate": 1.1470294058811765e-06, "loss": 0.0023, "step": 22142 }, { "epoch": 6.87, "learning_rate": 1.1466293258651731e-06, "loss": 0.0008, "step": 22143 }, { "epoch": 6.87, "learning_rate": 1.1462292458491698e-06, "loss": 0.0275, "step": 22144 }, { "epoch": 6.87, "learning_rate": 1.1458291658331667e-06, "loss": 0.0008, "step": 22145 }, { "epoch": 6.87, "learning_rate": 1.1454290858171634e-06, "loss": 0.0026, "step": 22146 }, { "epoch": 6.87, "learning_rate": 1.1450290058011603e-06, "loss": 0.0045, "step": 22147 }, { "epoch": 6.87, "learning_rate": 1.1446289257851572e-06, "loss": 0.0066, "step": 22148 }, { "epoch": 6.87, "learning_rate": 1.144228845769154e-06, "loss": 0.0022, "step": 22149 }, { "epoch": 6.87, "learning_rate": 1.1438287657531508e-06, "loss": 0.0016, "step": 22150 }, { "epoch": 6.88, "learning_rate": 1.1434286857371475e-06, "loss": 0.0021, "step": 22151 }, { "epoch": 6.88, "learning_rate": 1.1430286057211442e-06, "loss": 0.0015, "step": 22152 }, { "epoch": 6.88, "learning_rate": 1.142628525705141e-06, "loss": 0.0013, "step": 22153 }, { "epoch": 6.88, "learning_rate": 1.142228445689138e-06, "loss": 0.002, "step": 22154 }, { "epoch": 6.88, "learning_rate": 1.1418283656731346e-06, "loss": 0.0059, "step": 22155 }, { "epoch": 6.88, "learning_rate": 1.1414282856571315e-06, "loss": 0.0152, "step": 22156 }, { "epoch": 6.88, "learning_rate": 1.1410282056411282e-06, "loss": 0.0014, "step": 22157 }, { "epoch": 6.88, "learning_rate": 1.1406281256251251e-06, "loss": 0.0015, "step": 22158 }, { "epoch": 6.88, "learning_rate": 1.140228045609122e-06, "loss": 0.0068, "step": 22159 }, { "epoch": 6.88, "learning_rate": 1.1398279655931187e-06, "loss": 0.0065, "step": 22160 }, { "epoch": 6.88, "learning_rate": 1.1394278855771156e-06, "loss": 0.002, "step": 22161 }, { "epoch": 6.88, "learning_rate": 1.1390278055611123e-06, "loss": 0.0015, "step": 22162 }, { "epoch": 6.88, "learning_rate": 1.138627725545109e-06, "loss": 0.0009, "step": 22163 }, { "epoch": 6.88, "learning_rate": 1.1382276455291059e-06, "loss": 0.0013, "step": 22164 }, { "epoch": 6.88, "learning_rate": 1.1378275655131028e-06, "loss": 0.0008, "step": 22165 }, { "epoch": 6.88, "learning_rate": 1.1374274854970997e-06, "loss": 0.0024, "step": 22166 }, { "epoch": 6.88, "learning_rate": 1.1370274054810964e-06, "loss": 0.0015, "step": 22167 }, { "epoch": 6.88, "learning_rate": 1.136627325465093e-06, "loss": 0.0059, "step": 22168 }, { "epoch": 6.88, "learning_rate": 1.13622724544909e-06, "loss": 0.0017, "step": 22169 }, { "epoch": 6.88, "learning_rate": 1.1358271654330866e-06, "loss": 0.0031, "step": 22170 }, { "epoch": 6.88, "learning_rate": 1.1354270854170835e-06, "loss": 0.0015, "step": 22171 }, { "epoch": 6.88, "learning_rate": 1.1350270054010804e-06, "loss": 0.0054, "step": 22172 }, { "epoch": 6.88, "learning_rate": 1.1346269253850771e-06, "loss": 0.0015, "step": 22173 }, { "epoch": 6.88, "learning_rate": 1.1342268453690738e-06, "loss": 0.0019, "step": 22174 }, { "epoch": 6.88, "learning_rate": 1.1338267653530707e-06, "loss": 0.0016, "step": 22175 }, { "epoch": 6.88, "learning_rate": 1.1334266853370674e-06, "loss": 0.0011, "step": 22176 }, { "epoch": 6.88, "learning_rate": 1.1330266053210643e-06, "loss": 0.0032, "step": 22177 }, { "epoch": 6.88, "learning_rate": 1.1326265253050612e-06, "loss": 0.0034, "step": 22178 }, { "epoch": 6.88, "learning_rate": 1.1322264452890579e-06, "loss": 0.0101, "step": 22179 }, { "epoch": 6.88, "learning_rate": 1.1318263652730548e-06, "loss": 0.0016, "step": 22180 }, { "epoch": 6.88, "learning_rate": 1.1314262852570515e-06, "loss": 0.0028, "step": 22181 }, { "epoch": 6.88, "learning_rate": 1.1310262052410484e-06, "loss": 0.0015, "step": 22182 }, { "epoch": 6.89, "learning_rate": 1.130626125225045e-06, "loss": 0.0012, "step": 22183 }, { "epoch": 6.89, "learning_rate": 1.130226045209042e-06, "loss": 0.0024, "step": 22184 }, { "epoch": 6.89, "learning_rate": 1.1298259651930386e-06, "loss": 0.0014, "step": 22185 }, { "epoch": 6.89, "learning_rate": 1.1294258851770355e-06, "loss": 0.004, "step": 22186 }, { "epoch": 6.89, "learning_rate": 1.1290258051610322e-06, "loss": 0.0019, "step": 22187 }, { "epoch": 6.89, "learning_rate": 1.1286257251450291e-06, "loss": 0.0015, "step": 22188 }, { "epoch": 6.89, "learning_rate": 1.128225645129026e-06, "loss": 0.0018, "step": 22189 }, { "epoch": 6.89, "learning_rate": 1.1278255651130227e-06, "loss": 0.006, "step": 22190 }, { "epoch": 6.89, "learning_rate": 1.1274254850970196e-06, "loss": 0.001, "step": 22191 }, { "epoch": 6.89, "learning_rate": 1.1270254050810163e-06, "loss": 0.0015, "step": 22192 }, { "epoch": 6.89, "learning_rate": 1.126625325065013e-06, "loss": 0.0014, "step": 22193 }, { "epoch": 6.89, "learning_rate": 1.1262252450490099e-06, "loss": 0.0012, "step": 22194 }, { "epoch": 6.89, "learning_rate": 1.1258251650330068e-06, "loss": 0.0019, "step": 22195 }, { "epoch": 6.89, "learning_rate": 1.1254250850170034e-06, "loss": 0.0008, "step": 22196 }, { "epoch": 6.89, "learning_rate": 1.1250250050010003e-06, "loss": 0.0014, "step": 22197 }, { "epoch": 6.89, "learning_rate": 1.124624924984997e-06, "loss": 0.0032, "step": 22198 }, { "epoch": 6.89, "learning_rate": 1.124224844968994e-06, "loss": 0.0012, "step": 22199 }, { "epoch": 6.89, "learning_rate": 1.1238247649529906e-06, "loss": 0.0344, "step": 22200 }, { "epoch": 6.89, "learning_rate": 1.1234246849369875e-06, "loss": 0.0115, "step": 22201 }, { "epoch": 6.89, "learning_rate": 1.1230246049209844e-06, "loss": 0.0026, "step": 22202 }, { "epoch": 6.89, "learning_rate": 1.122624524904981e-06, "loss": 0.0324, "step": 22203 }, { "epoch": 6.89, "learning_rate": 1.1222244448889778e-06, "loss": 0.0031, "step": 22204 }, { "epoch": 6.89, "learning_rate": 1.1218243648729747e-06, "loss": 0.0047, "step": 22205 }, { "epoch": 6.89, "learning_rate": 1.1214242848569714e-06, "loss": 0.0036, "step": 22206 }, { "epoch": 6.89, "learning_rate": 1.1210242048409683e-06, "loss": 0.0008, "step": 22207 }, { "epoch": 6.89, "learning_rate": 1.1206241248249652e-06, "loss": 0.0004, "step": 22208 }, { "epoch": 6.89, "learning_rate": 1.1202240448089618e-06, "loss": 0.0061, "step": 22209 }, { "epoch": 6.89, "learning_rate": 1.1198239647929587e-06, "loss": 0.0011, "step": 22210 }, { "epoch": 6.89, "learning_rate": 1.1194238847769554e-06, "loss": 0.0049, "step": 22211 }, { "epoch": 6.89, "learning_rate": 1.1190238047609523e-06, "loss": 0.0149, "step": 22212 }, { "epoch": 6.89, "learning_rate": 1.1186237247449492e-06, "loss": 0.0019, "step": 22213 }, { "epoch": 6.89, "learning_rate": 1.118223644728946e-06, "loss": 0.0039, "step": 22214 }, { "epoch": 6.9, "learning_rate": 1.1178235647129426e-06, "loss": 0.0057, "step": 22215 }, { "epoch": 6.9, "learning_rate": 1.1174234846969395e-06, "loss": 0.0035, "step": 22216 }, { "epoch": 6.9, "learning_rate": 1.1170234046809362e-06, "loss": 0.0061, "step": 22217 }, { "epoch": 6.9, "learning_rate": 1.116623324664933e-06, "loss": 0.003, "step": 22218 }, { "epoch": 6.9, "learning_rate": 1.11622324464893e-06, "loss": 0.0036, "step": 22219 }, { "epoch": 6.9, "learning_rate": 1.1158231646329267e-06, "loss": 0.001, "step": 22220 }, { "epoch": 6.9, "learning_rate": 1.1154230846169236e-06, "loss": 0.0038, "step": 22221 }, { "epoch": 6.9, "learning_rate": 1.1150230046009203e-06, "loss": 0.001, "step": 22222 }, { "epoch": 6.9, "learning_rate": 1.114622924584917e-06, "loss": 0.0014, "step": 22223 }, { "epoch": 6.9, "learning_rate": 1.1142228445689138e-06, "loss": 0.0018, "step": 22224 }, { "epoch": 6.9, "learning_rate": 1.1138227645529107e-06, "loss": 0.0022, "step": 22225 }, { "epoch": 6.9, "learning_rate": 1.1134226845369074e-06, "loss": 0.0005, "step": 22226 }, { "epoch": 6.9, "learning_rate": 1.1130226045209043e-06, "loss": 0.0017, "step": 22227 }, { "epoch": 6.9, "learning_rate": 1.112622524504901e-06, "loss": 0.0013, "step": 22228 }, { "epoch": 6.9, "learning_rate": 1.112222444488898e-06, "loss": 0.0011, "step": 22229 }, { "epoch": 6.9, "learning_rate": 1.1118223644728946e-06, "loss": 0.0015, "step": 22230 }, { "epoch": 6.9, "learning_rate": 1.1114222844568915e-06, "loss": 0.001, "step": 22231 }, { "epoch": 6.9, "learning_rate": 1.1110222044408884e-06, "loss": 0.005, "step": 22232 }, { "epoch": 6.9, "learning_rate": 1.110622124424885e-06, "loss": 0.0013, "step": 22233 }, { "epoch": 6.9, "learning_rate": 1.1102220444088818e-06, "loss": 0.0019, "step": 22234 }, { "epoch": 6.9, "learning_rate": 1.1098219643928787e-06, "loss": 0.0024, "step": 22235 }, { "epoch": 6.9, "learning_rate": 1.1094218843768756e-06, "loss": 0.003, "step": 22236 }, { "epoch": 6.9, "learning_rate": 1.1090218043608722e-06, "loss": 0.0089, "step": 22237 }, { "epoch": 6.9, "learning_rate": 1.1086217243448691e-06, "loss": 0.0027, "step": 22238 }, { "epoch": 6.9, "learning_rate": 1.1082216443288658e-06, "loss": 0.0012, "step": 22239 }, { "epoch": 6.9, "learning_rate": 1.1078215643128627e-06, "loss": 0.0014, "step": 22240 }, { "epoch": 6.9, "learning_rate": 1.1074214842968594e-06, "loss": 0.0008, "step": 22241 }, { "epoch": 6.9, "learning_rate": 1.1070214042808563e-06, "loss": 0.0014, "step": 22242 }, { "epoch": 6.9, "learning_rate": 1.1066213242648532e-06, "loss": 0.0016, "step": 22243 }, { "epoch": 6.9, "learning_rate": 1.1062212442488499e-06, "loss": 0.0034, "step": 22244 }, { "epoch": 6.9, "learning_rate": 1.1058211642328466e-06, "loss": 0.0012, "step": 22245 }, { "epoch": 6.9, "learning_rate": 1.1054210842168435e-06, "loss": 0.0018, "step": 22246 }, { "epoch": 6.9, "learning_rate": 1.1050210042008402e-06, "loss": 0.0034, "step": 22247 }, { "epoch": 6.91, "learning_rate": 1.104620924184837e-06, "loss": 0.0257, "step": 22248 }, { "epoch": 6.91, "learning_rate": 1.104220844168834e-06, "loss": 0.0008, "step": 22249 }, { "epoch": 6.91, "learning_rate": 1.1038207641528306e-06, "loss": 0.0035, "step": 22250 }, { "epoch": 6.91, "learning_rate": 1.1034206841368275e-06, "loss": 0.0026, "step": 22251 }, { "epoch": 6.91, "learning_rate": 1.1030206041208242e-06, "loss": 0.0009, "step": 22252 }, { "epoch": 6.91, "learning_rate": 1.102620524104821e-06, "loss": 0.0015, "step": 22253 }, { "epoch": 6.91, "learning_rate": 1.1022204440888178e-06, "loss": 0.0018, "step": 22254 }, { "epoch": 6.91, "learning_rate": 1.1018203640728147e-06, "loss": 0.0016, "step": 22255 }, { "epoch": 6.91, "learning_rate": 1.1014202840568114e-06, "loss": 0.0024, "step": 22256 }, { "epoch": 6.91, "learning_rate": 1.1010202040408083e-06, "loss": 0.0007, "step": 22257 }, { "epoch": 6.91, "learning_rate": 1.100620124024805e-06, "loss": 0.0031, "step": 22258 }, { "epoch": 6.91, "learning_rate": 1.1002200440088019e-06, "loss": 0.0011, "step": 22259 }, { "epoch": 6.91, "learning_rate": 1.0998199639927986e-06, "loss": 0.0033, "step": 22260 }, { "epoch": 6.91, "learning_rate": 1.0994198839767955e-06, "loss": 0.0008, "step": 22261 }, { "epoch": 6.91, "learning_rate": 1.0990198039607924e-06, "loss": 0.0025, "step": 22262 }, { "epoch": 6.91, "learning_rate": 1.098619723944789e-06, "loss": 0.0025, "step": 22263 }, { "epoch": 6.91, "learning_rate": 1.0982196439287857e-06, "loss": 0.0016, "step": 22264 }, { "epoch": 6.91, "learning_rate": 1.0978195639127826e-06, "loss": 0.0032, "step": 22265 }, { "epoch": 6.91, "learning_rate": 1.0974194838967795e-06, "loss": 0.0015, "step": 22266 }, { "epoch": 6.91, "learning_rate": 1.0970194038807762e-06, "loss": 0.0026, "step": 22267 }, { "epoch": 6.91, "learning_rate": 1.0966193238647731e-06, "loss": 0.0013, "step": 22268 }, { "epoch": 6.91, "learning_rate": 1.0962192438487698e-06, "loss": 0.0012, "step": 22269 }, { "epoch": 6.91, "learning_rate": 1.0958191638327667e-06, "loss": 0.0211, "step": 22270 }, { "epoch": 6.91, "learning_rate": 1.0954190838167634e-06, "loss": 0.0017, "step": 22271 }, { "epoch": 6.91, "learning_rate": 1.0950190038007603e-06, "loss": 0.0239, "step": 22272 }, { "epoch": 6.91, "learning_rate": 1.0946189237847572e-06, "loss": 0.001, "step": 22273 }, { "epoch": 6.91, "learning_rate": 1.0942188437687539e-06, "loss": 0.0011, "step": 22274 }, { "epoch": 6.91, "learning_rate": 1.0938187637527506e-06, "loss": 0.0014, "step": 22275 }, { "epoch": 6.91, "learning_rate": 1.0934186837367475e-06, "loss": 0.0352, "step": 22276 }, { "epoch": 6.91, "learning_rate": 1.0930186037207441e-06, "loss": 0.0007, "step": 22277 }, { "epoch": 6.91, "learning_rate": 1.092618523704741e-06, "loss": 0.0061, "step": 22278 }, { "epoch": 6.91, "learning_rate": 1.092218443688738e-06, "loss": 0.0007, "step": 22279 }, { "epoch": 6.92, "learning_rate": 1.0918183636727346e-06, "loss": 0.0011, "step": 22280 }, { "epoch": 6.92, "learning_rate": 1.0914182836567315e-06, "loss": 0.0021, "step": 22281 }, { "epoch": 6.92, "learning_rate": 1.0910182036407282e-06, "loss": 0.0062, "step": 22282 }, { "epoch": 6.92, "learning_rate": 1.0906181236247249e-06, "loss": 0.0038, "step": 22283 }, { "epoch": 6.92, "learning_rate": 1.0902180436087218e-06, "loss": 0.06, "step": 22284 }, { "epoch": 6.92, "learning_rate": 1.0898179635927187e-06, "loss": 0.0036, "step": 22285 }, { "epoch": 6.92, "learning_rate": 1.0894178835767154e-06, "loss": 0.0011, "step": 22286 }, { "epoch": 6.92, "learning_rate": 1.0890178035607123e-06, "loss": 0.001, "step": 22287 }, { "epoch": 6.92, "learning_rate": 1.088617723544709e-06, "loss": 0.0036, "step": 22288 }, { "epoch": 6.92, "learning_rate": 1.0882176435287059e-06, "loss": 0.0014, "step": 22289 }, { "epoch": 6.92, "learning_rate": 1.0878175635127028e-06, "loss": 0.0041, "step": 22290 }, { "epoch": 6.92, "learning_rate": 1.0874174834966994e-06, "loss": 0.0012, "step": 22291 }, { "epoch": 6.92, "learning_rate": 1.0870174034806963e-06, "loss": 0.0014, "step": 22292 }, { "epoch": 6.92, "learning_rate": 1.086617323464693e-06, "loss": 0.0014, "step": 22293 }, { "epoch": 6.92, "learning_rate": 1.0862172434486897e-06, "loss": 0.0021, "step": 22294 }, { "epoch": 6.92, "learning_rate": 1.0858171634326866e-06, "loss": 0.0057, "step": 22295 }, { "epoch": 6.92, "learning_rate": 1.0854170834166835e-06, "loss": 0.0013, "step": 22296 }, { "epoch": 6.92, "learning_rate": 1.0850170034006802e-06, "loss": 0.0084, "step": 22297 }, { "epoch": 6.92, "learning_rate": 1.084616923384677e-06, "loss": 0.0008, "step": 22298 }, { "epoch": 6.92, "learning_rate": 1.0842168433686738e-06, "loss": 0.0009, "step": 22299 }, { "epoch": 6.92, "learning_rate": 1.0838167633526705e-06, "loss": 0.0248, "step": 22300 }, { "epoch": 6.92, "learning_rate": 1.0834166833366674e-06, "loss": 0.0013, "step": 22301 }, { "epoch": 6.92, "learning_rate": 1.0830166033206643e-06, "loss": 0.0012, "step": 22302 }, { "epoch": 6.92, "learning_rate": 1.0826165233046612e-06, "loss": 0.0275, "step": 22303 }, { "epoch": 6.92, "learning_rate": 1.0822164432886578e-06, "loss": 0.0012, "step": 22304 }, { "epoch": 6.92, "learning_rate": 1.0818163632726545e-06, "loss": 0.0166, "step": 22305 }, { "epoch": 6.92, "learning_rate": 1.0814162832566514e-06, "loss": 0.0014, "step": 22306 }, { "epoch": 6.92, "learning_rate": 1.0810162032406481e-06, "loss": 0.0037, "step": 22307 }, { "epoch": 6.92, "learning_rate": 1.080616123224645e-06, "loss": 0.0013, "step": 22308 }, { "epoch": 6.92, "learning_rate": 1.080216043208642e-06, "loss": 0.0017, "step": 22309 }, { "epoch": 6.92, "learning_rate": 1.0798159631926386e-06, "loss": 0.0011, "step": 22310 }, { "epoch": 6.92, "learning_rate": 1.0794158831766355e-06, "loss": 0.0009, "step": 22311 }, { "epoch": 6.93, "learning_rate": 1.0790158031606322e-06, "loss": 0.0009, "step": 22312 }, { "epoch": 6.93, "learning_rate": 1.078615723144629e-06, "loss": 0.0018, "step": 22313 }, { "epoch": 6.93, "learning_rate": 1.0782156431286258e-06, "loss": 0.0015, "step": 22314 }, { "epoch": 6.93, "learning_rate": 1.0778155631126227e-06, "loss": 0.0055, "step": 22315 }, { "epoch": 6.93, "learning_rate": 1.0774154830966193e-06, "loss": 0.0008, "step": 22316 }, { "epoch": 6.93, "learning_rate": 1.0770154030806162e-06, "loss": 0.0273, "step": 22317 }, { "epoch": 6.93, "learning_rate": 1.076615323064613e-06, "loss": 0.0794, "step": 22318 }, { "epoch": 6.93, "learning_rate": 1.0762152430486098e-06, "loss": 0.0009, "step": 22319 }, { "epoch": 6.93, "learning_rate": 1.0758151630326067e-06, "loss": 0.0021, "step": 22320 }, { "epoch": 6.93, "learning_rate": 1.0754150830166034e-06, "loss": 0.0008, "step": 22321 }, { "epoch": 6.93, "learning_rate": 1.0750150030006003e-06, "loss": 0.0028, "step": 22322 }, { "epoch": 6.93, "learning_rate": 1.074614922984597e-06, "loss": 0.0021, "step": 22323 }, { "epoch": 6.93, "learning_rate": 1.0742148429685937e-06, "loss": 0.0023, "step": 22324 }, { "epoch": 6.93, "learning_rate": 1.0738147629525906e-06, "loss": 0.0007, "step": 22325 }, { "epoch": 6.93, "learning_rate": 1.0734146829365875e-06, "loss": 0.0038, "step": 22326 }, { "epoch": 6.93, "learning_rate": 1.0730146029205842e-06, "loss": 0.0076, "step": 22327 }, { "epoch": 6.93, "learning_rate": 1.072614522904581e-06, "loss": 0.0014, "step": 22328 }, { "epoch": 6.93, "learning_rate": 1.0722144428885778e-06, "loss": 0.0018, "step": 22329 }, { "epoch": 6.93, "learning_rate": 1.0718143628725744e-06, "loss": 0.0013, "step": 22330 }, { "epoch": 6.93, "learning_rate": 1.0714142828565713e-06, "loss": 0.0028, "step": 22331 }, { "epoch": 6.93, "learning_rate": 1.0710142028405682e-06, "loss": 0.002, "step": 22332 }, { "epoch": 6.93, "learning_rate": 1.0706141228245651e-06, "loss": 0.0016, "step": 22333 }, { "epoch": 6.93, "learning_rate": 1.0702140428085618e-06, "loss": 0.0044, "step": 22334 }, { "epoch": 6.93, "learning_rate": 1.0698139627925585e-06, "loss": 0.0027, "step": 22335 }, { "epoch": 6.93, "learning_rate": 1.0694138827765554e-06, "loss": 0.0011, "step": 22336 }, { "epoch": 6.93, "learning_rate": 1.069013802760552e-06, "loss": 0.0013, "step": 22337 }, { "epoch": 6.93, "learning_rate": 1.068613722744549e-06, "loss": 0.0005, "step": 22338 }, { "epoch": 6.93, "learning_rate": 1.0682136427285459e-06, "loss": 0.0015, "step": 22339 }, { "epoch": 6.93, "learning_rate": 1.0678135627125426e-06, "loss": 0.003, "step": 22340 }, { "epoch": 6.93, "learning_rate": 1.0674134826965393e-06, "loss": 0.0025, "step": 22341 }, { "epoch": 6.93, "learning_rate": 1.0670134026805362e-06, "loss": 0.0009, "step": 22342 }, { "epoch": 6.93, "learning_rate": 1.066613322664533e-06, "loss": 0.0016, "step": 22343 }, { "epoch": 6.94, "learning_rate": 1.06621324264853e-06, "loss": 0.0016, "step": 22344 }, { "epoch": 6.94, "learning_rate": 1.0658131626325266e-06, "loss": 0.0008, "step": 22345 }, { "epoch": 6.94, "learning_rate": 1.0654130826165233e-06, "loss": 0.0014, "step": 22346 }, { "epoch": 6.94, "learning_rate": 1.0650130026005202e-06, "loss": 0.0028, "step": 22347 }, { "epoch": 6.94, "learning_rate": 1.064612922584517e-06, "loss": 0.0024, "step": 22348 }, { "epoch": 6.94, "learning_rate": 1.0642128425685138e-06, "loss": 0.0056, "step": 22349 }, { "epoch": 6.94, "learning_rate": 1.0638127625525107e-06, "loss": 0.0046, "step": 22350 }, { "epoch": 6.94, "learning_rate": 1.0634126825365074e-06, "loss": 0.0004, "step": 22351 }, { "epoch": 6.94, "learning_rate": 1.0630126025205043e-06, "loss": 0.0012, "step": 22352 }, { "epoch": 6.94, "learning_rate": 1.062612522504501e-06, "loss": 0.0031, "step": 22353 }, { "epoch": 6.94, "learning_rate": 1.0622124424884977e-06, "loss": 0.0009, "step": 22354 }, { "epoch": 6.94, "learning_rate": 1.0618123624724946e-06, "loss": 0.001, "step": 22355 }, { "epoch": 6.94, "learning_rate": 1.0614122824564915e-06, "loss": 0.0019, "step": 22356 }, { "epoch": 6.94, "learning_rate": 1.0610122024404881e-06, "loss": 0.004, "step": 22357 }, { "epoch": 6.94, "learning_rate": 1.060612122424485e-06, "loss": 0.0036, "step": 22358 }, { "epoch": 6.94, "learning_rate": 1.0602120424084817e-06, "loss": 0.0007, "step": 22359 }, { "epoch": 6.94, "learning_rate": 1.0598119623924784e-06, "loss": 0.0014, "step": 22360 }, { "epoch": 6.94, "learning_rate": 1.0594118823764753e-06, "loss": 0.0034, "step": 22361 }, { "epoch": 6.94, "learning_rate": 1.0590118023604722e-06, "loss": 0.0011, "step": 22362 }, { "epoch": 6.94, "learning_rate": 1.0586117223444691e-06, "loss": 0.0022, "step": 22363 }, { "epoch": 6.94, "learning_rate": 1.0582116423284658e-06, "loss": 0.0025, "step": 22364 }, { "epoch": 6.94, "learning_rate": 1.0578115623124625e-06, "loss": 0.0037, "step": 22365 }, { "epoch": 6.94, "learning_rate": 1.0574114822964594e-06, "loss": 0.0013, "step": 22366 }, { "epoch": 6.94, "learning_rate": 1.0570114022804563e-06, "loss": 0.0012, "step": 22367 }, { "epoch": 6.94, "learning_rate": 1.056611322264453e-06, "loss": 0.0014, "step": 22368 }, { "epoch": 6.94, "learning_rate": 1.0562112422484499e-06, "loss": 0.0058, "step": 22369 }, { "epoch": 6.94, "learning_rate": 1.0558111622324465e-06, "loss": 0.0011, "step": 22370 }, { "epoch": 6.94, "learning_rate": 1.0554110822164432e-06, "loss": 0.0022, "step": 22371 }, { "epoch": 6.94, "learning_rate": 1.0550110022004401e-06, "loss": 0.0024, "step": 22372 }, { "epoch": 6.94, "learning_rate": 1.054610922184437e-06, "loss": 0.0046, "step": 22373 }, { "epoch": 6.94, "learning_rate": 1.054210842168434e-06, "loss": 0.0103, "step": 22374 }, { "epoch": 6.94, "learning_rate": 1.0538107621524306e-06, "loss": 0.001, "step": 22375 }, { "epoch": 6.95, "learning_rate": 1.0534106821364273e-06, "loss": 0.0016, "step": 22376 }, { "epoch": 6.95, "learning_rate": 1.0530106021204242e-06, "loss": 0.0024, "step": 22377 }, { "epoch": 6.95, "learning_rate": 1.0526105221044209e-06, "loss": 0.0094, "step": 22378 }, { "epoch": 6.95, "learning_rate": 1.0522104420884178e-06, "loss": 0.0017, "step": 22379 }, { "epoch": 6.95, "learning_rate": 1.0518103620724147e-06, "loss": 0.0008, "step": 22380 }, { "epoch": 6.95, "learning_rate": 1.0514102820564114e-06, "loss": 0.0003, "step": 22381 }, { "epoch": 6.95, "learning_rate": 1.051010202040408e-06, "loss": 0.0031, "step": 22382 }, { "epoch": 6.95, "learning_rate": 1.050610122024405e-06, "loss": 0.0024, "step": 22383 }, { "epoch": 6.95, "learning_rate": 1.0502100420084016e-06, "loss": 0.003, "step": 22384 }, { "epoch": 6.95, "learning_rate": 1.0498099619923985e-06, "loss": 0.0233, "step": 22385 }, { "epoch": 6.95, "learning_rate": 1.0494098819763954e-06, "loss": 0.0014, "step": 22386 }, { "epoch": 6.95, "learning_rate": 1.0490098019603921e-06, "loss": 0.0091, "step": 22387 }, { "epoch": 6.95, "learning_rate": 1.048609721944389e-06, "loss": 0.0026, "step": 22388 }, { "epoch": 6.95, "learning_rate": 1.0482096419283857e-06, "loss": 0.001, "step": 22389 }, { "epoch": 6.95, "learning_rate": 1.0478095619123826e-06, "loss": 0.0054, "step": 22390 }, { "epoch": 6.95, "learning_rate": 1.0474094818963793e-06, "loss": 0.003, "step": 22391 }, { "epoch": 6.95, "learning_rate": 1.0470094018803762e-06, "loss": 0.0034, "step": 22392 }, { "epoch": 6.95, "learning_rate": 1.046609321864373e-06, "loss": 0.0014, "step": 22393 }, { "epoch": 6.95, "learning_rate": 1.0462092418483698e-06, "loss": 0.0011, "step": 22394 }, { "epoch": 6.95, "learning_rate": 1.0458091618323665e-06, "loss": 0.0386, "step": 22395 }, { "epoch": 6.95, "learning_rate": 1.0454090818163634e-06, "loss": 0.0029, "step": 22396 }, { "epoch": 6.95, "learning_rate": 1.0450090018003603e-06, "loss": 0.0015, "step": 22397 }, { "epoch": 6.95, "learning_rate": 1.044608921784357e-06, "loss": 0.0096, "step": 22398 }, { "epoch": 6.95, "learning_rate": 1.0442088417683538e-06, "loss": 0.0014, "step": 22399 }, { "epoch": 6.95, "learning_rate": 1.0438087617523505e-06, "loss": 0.0012, "step": 22400 }, { "epoch": 6.95, "learning_rate": 1.0434086817363472e-06, "loss": 0.0022, "step": 22401 }, { "epoch": 6.95, "learning_rate": 1.0430086017203441e-06, "loss": 0.003, "step": 22402 }, { "epoch": 6.95, "learning_rate": 1.042608521704341e-06, "loss": 0.0015, "step": 22403 }, { "epoch": 6.95, "learning_rate": 1.042208441688338e-06, "loss": 0.0008, "step": 22404 }, { "epoch": 6.95, "learning_rate": 1.0418083616723346e-06, "loss": 0.002, "step": 22405 }, { "epoch": 6.95, "learning_rate": 1.0414082816563313e-06, "loss": 0.0051, "step": 22406 }, { "epoch": 6.95, "learning_rate": 1.0410082016403282e-06, "loss": 0.0006, "step": 22407 }, { "epoch": 6.95, "learning_rate": 1.0406081216243249e-06, "loss": 0.0008, "step": 22408 }, { "epoch": 6.96, "learning_rate": 1.0402080416083218e-06, "loss": 0.0126, "step": 22409 }, { "epoch": 6.96, "learning_rate": 1.0398079615923187e-06, "loss": 0.0098, "step": 22410 }, { "epoch": 6.96, "learning_rate": 1.0394078815763153e-06, "loss": 0.0012, "step": 22411 }, { "epoch": 6.96, "learning_rate": 1.039007801560312e-06, "loss": 0.0009, "step": 22412 }, { "epoch": 6.96, "learning_rate": 1.038607721544309e-06, "loss": 0.0013, "step": 22413 }, { "epoch": 6.96, "learning_rate": 1.0382076415283056e-06, "loss": 0.0027, "step": 22414 }, { "epoch": 6.96, "learning_rate": 1.0378075615123025e-06, "loss": 0.0042, "step": 22415 }, { "epoch": 6.96, "learning_rate": 1.0374074814962994e-06, "loss": 0.0224, "step": 22416 }, { "epoch": 6.96, "learning_rate": 1.037007401480296e-06, "loss": 0.0029, "step": 22417 }, { "epoch": 6.96, "learning_rate": 1.036607321464293e-06, "loss": 0.0064, "step": 22418 }, { "epoch": 6.96, "learning_rate": 1.0362072414482897e-06, "loss": 0.0015, "step": 22419 }, { "epoch": 6.96, "learning_rate": 1.0358071614322866e-06, "loss": 0.0019, "step": 22420 }, { "epoch": 6.96, "learning_rate": 1.0354070814162835e-06, "loss": 0.0021, "step": 22421 }, { "epoch": 6.96, "learning_rate": 1.0350070014002802e-06, "loss": 0.0023, "step": 22422 }, { "epoch": 6.96, "learning_rate": 1.0346069213842769e-06, "loss": 0.0012, "step": 22423 }, { "epoch": 6.96, "learning_rate": 1.0342068413682737e-06, "loss": 0.0053, "step": 22424 }, { "epoch": 6.96, "learning_rate": 1.0338067613522704e-06, "loss": 0.0007, "step": 22425 }, { "epoch": 6.96, "learning_rate": 1.0334066813362673e-06, "loss": 0.0049, "step": 22426 }, { "epoch": 6.96, "learning_rate": 1.0330066013202642e-06, "loss": 0.001, "step": 22427 }, { "epoch": 6.96, "learning_rate": 1.032606521304261e-06, "loss": 0.0012, "step": 22428 }, { "epoch": 6.96, "learning_rate": 1.0322064412882578e-06, "loss": 0.0019, "step": 22429 }, { "epoch": 6.96, "learning_rate": 1.0318063612722545e-06, "loss": 0.036, "step": 22430 }, { "epoch": 6.96, "learning_rate": 1.0314062812562512e-06, "loss": 0.0015, "step": 22431 }, { "epoch": 6.96, "learning_rate": 1.031006201240248e-06, "loss": 0.0063, "step": 22432 }, { "epoch": 6.96, "learning_rate": 1.030606121224245e-06, "loss": 0.005, "step": 22433 }, { "epoch": 6.96, "learning_rate": 1.0302060412082417e-06, "loss": 0.0034, "step": 22434 }, { "epoch": 6.96, "learning_rate": 1.0298059611922386e-06, "loss": 0.0013, "step": 22435 }, { "epoch": 6.96, "learning_rate": 1.0294058811762353e-06, "loss": 0.0028, "step": 22436 }, { "epoch": 6.96, "learning_rate": 1.0290058011602322e-06, "loss": 0.005, "step": 22437 }, { "epoch": 6.96, "learning_rate": 1.0286057211442288e-06, "loss": 0.0014, "step": 22438 }, { "epoch": 6.96, "learning_rate": 1.0282056411282257e-06, "loss": 0.004, "step": 22439 }, { "epoch": 6.96, "learning_rate": 1.0278055611122226e-06, "loss": 0.0652, "step": 22440 }, { "epoch": 6.97, "learning_rate": 1.0274054810962193e-06, "loss": 0.0009, "step": 22441 }, { "epoch": 6.97, "learning_rate": 1.027005401080216e-06, "loss": 0.0005, "step": 22442 }, { "epoch": 6.97, "learning_rate": 1.026605321064213e-06, "loss": 0.0014, "step": 22443 }, { "epoch": 6.97, "learning_rate": 1.0262052410482098e-06, "loss": 0.0027, "step": 22444 }, { "epoch": 6.97, "learning_rate": 1.0258051610322065e-06, "loss": 0.0355, "step": 22445 }, { "epoch": 6.97, "learning_rate": 1.0254050810162034e-06, "loss": 0.0012, "step": 22446 }, { "epoch": 6.97, "learning_rate": 1.0250050010002e-06, "loss": 0.0025, "step": 22447 }, { "epoch": 6.97, "learning_rate": 1.024604920984197e-06, "loss": 0.0557, "step": 22448 }, { "epoch": 6.97, "learning_rate": 1.0242048409681937e-06, "loss": 0.0015, "step": 22449 }, { "epoch": 6.97, "learning_rate": 1.0238047609521906e-06, "loss": 0.003, "step": 22450 }, { "epoch": 6.97, "learning_rate": 1.0234046809361875e-06, "loss": 0.0017, "step": 22451 }, { "epoch": 6.97, "learning_rate": 1.0230046009201841e-06, "loss": 0.0019, "step": 22452 }, { "epoch": 6.97, "learning_rate": 1.0226045209041808e-06, "loss": 0.007, "step": 22453 }, { "epoch": 6.97, "learning_rate": 1.0222044408881777e-06, "loss": 0.0012, "step": 22454 }, { "epoch": 6.97, "learning_rate": 1.0218043608721744e-06, "loss": 0.0048, "step": 22455 }, { "epoch": 6.97, "learning_rate": 1.0214042808561713e-06, "loss": 0.0037, "step": 22456 }, { "epoch": 6.97, "learning_rate": 1.0210042008401682e-06, "loss": 0.0017, "step": 22457 }, { "epoch": 6.97, "learning_rate": 1.0206041208241649e-06, "loss": 0.0421, "step": 22458 }, { "epoch": 6.97, "learning_rate": 1.0202040408081618e-06, "loss": 0.0029, "step": 22459 }, { "epoch": 6.97, "learning_rate": 1.0198039607921585e-06, "loss": 0.0011, "step": 22460 }, { "epoch": 6.97, "learning_rate": 1.0194038807761552e-06, "loss": 0.0011, "step": 22461 }, { "epoch": 6.97, "learning_rate": 1.019003800760152e-06, "loss": 0.0014, "step": 22462 }, { "epoch": 6.97, "learning_rate": 1.018603720744149e-06, "loss": 0.0044, "step": 22463 }, { "epoch": 6.97, "learning_rate": 1.0182036407281456e-06, "loss": 0.0009, "step": 22464 }, { "epoch": 6.97, "learning_rate": 1.0178035607121425e-06, "loss": 0.0082, "step": 22465 }, { "epoch": 6.97, "learning_rate": 1.0174034806961392e-06, "loss": 0.0016, "step": 22466 }, { "epoch": 6.97, "learning_rate": 1.0170034006801361e-06, "loss": 0.0008, "step": 22467 }, { "epoch": 6.97, "learning_rate": 1.0166033206641328e-06, "loss": 0.0025, "step": 22468 }, { "epoch": 6.97, "learning_rate": 1.0162032406481297e-06, "loss": 0.0018, "step": 22469 }, { "epoch": 6.97, "learning_rate": 1.0158031606321266e-06, "loss": 0.0079, "step": 22470 }, { "epoch": 6.97, "learning_rate": 1.0154030806161233e-06, "loss": 0.002, "step": 22471 }, { "epoch": 6.97, "learning_rate": 1.01500300060012e-06, "loss": 0.001, "step": 22472 }, { "epoch": 6.98, "learning_rate": 1.0146029205841169e-06, "loss": 0.0018, "step": 22473 }, { "epoch": 6.98, "learning_rate": 1.0142028405681138e-06, "loss": 0.0035, "step": 22474 }, { "epoch": 6.98, "learning_rate": 1.0138027605521105e-06, "loss": 0.0012, "step": 22475 }, { "epoch": 6.98, "learning_rate": 1.0134026805361074e-06, "loss": 0.0012, "step": 22476 }, { "epoch": 6.98, "learning_rate": 1.013002600520104e-06, "loss": 0.0009, "step": 22477 }, { "epoch": 6.98, "learning_rate": 1.012602520504101e-06, "loss": 0.0021, "step": 22478 }, { "epoch": 6.98, "learning_rate": 1.0122024404880976e-06, "loss": 0.0024, "step": 22479 }, { "epoch": 6.98, "learning_rate": 1.0118023604720945e-06, "loss": 0.0033, "step": 22480 }, { "epoch": 6.98, "learning_rate": 1.0114022804560914e-06, "loss": 0.0083, "step": 22481 }, { "epoch": 6.98, "learning_rate": 1.0110022004400881e-06, "loss": 0.0008, "step": 22482 }, { "epoch": 6.98, "learning_rate": 1.0106021204240848e-06, "loss": 0.0063, "step": 22483 }, { "epoch": 6.98, "learning_rate": 1.0102020404080817e-06, "loss": 0.0016, "step": 22484 }, { "epoch": 6.98, "learning_rate": 1.0098019603920784e-06, "loss": 0.0011, "step": 22485 }, { "epoch": 6.98, "learning_rate": 1.0094018803760753e-06, "loss": 0.0044, "step": 22486 }, { "epoch": 6.98, "learning_rate": 1.0090018003600722e-06, "loss": 0.0034, "step": 22487 }, { "epoch": 6.98, "learning_rate": 1.0086017203440689e-06, "loss": 0.0011, "step": 22488 }, { "epoch": 6.98, "learning_rate": 1.0082016403280658e-06, "loss": 0.0146, "step": 22489 }, { "epoch": 6.98, "learning_rate": 1.0078015603120625e-06, "loss": 0.0012, "step": 22490 }, { "epoch": 6.98, "learning_rate": 1.0074014802960594e-06, "loss": 0.0017, "step": 22491 }, { "epoch": 6.98, "learning_rate": 1.007001400280056e-06, "loss": 0.001, "step": 22492 }, { "epoch": 6.98, "learning_rate": 1.006601320264053e-06, "loss": 0.002, "step": 22493 }, { "epoch": 6.98, "learning_rate": 1.0062012402480496e-06, "loss": 0.0014, "step": 22494 }, { "epoch": 6.98, "learning_rate": 1.0058011602320465e-06, "loss": 0.0019, "step": 22495 }, { "epoch": 6.98, "learning_rate": 1.0054010802160432e-06, "loss": 0.0009, "step": 22496 }, { "epoch": 6.98, "learning_rate": 1.00500100020004e-06, "loss": 0.0021, "step": 22497 }, { "epoch": 6.98, "learning_rate": 1.004600920184037e-06, "loss": 0.003, "step": 22498 }, { "epoch": 6.98, "learning_rate": 1.0042008401680337e-06, "loss": 0.0023, "step": 22499 }, { "epoch": 6.98, "learning_rate": 1.0038007601520306e-06, "loss": 0.0006, "step": 22500 }, { "epoch": 6.98, "learning_rate": 1.0034006801360273e-06, "loss": 0.0007, "step": 22501 }, { "epoch": 6.98, "learning_rate": 1.003000600120024e-06, "loss": 0.0429, "step": 22502 }, { "epoch": 6.98, "learning_rate": 1.0026005201040209e-06, "loss": 0.0014, "step": 22503 }, { "epoch": 6.98, "learning_rate": 1.0022004400880178e-06, "loss": 0.0041, "step": 22504 }, { "epoch": 6.99, "learning_rate": 1.0018003600720144e-06, "loss": 0.0035, "step": 22505 }, { "epoch": 6.99, "learning_rate": 1.0014002800560113e-06, "loss": 0.0015, "step": 22506 }, { "epoch": 6.99, "learning_rate": 1.001000200040008e-06, "loss": 0.0417, "step": 22507 }, { "epoch": 6.99, "learning_rate": 1.000600120024005e-06, "loss": 0.0013, "step": 22508 }, { "epoch": 6.99, "learning_rate": 1.0002000400080016e-06, "loss": 0.0035, "step": 22509 }, { "epoch": 6.99, "learning_rate": 9.997999599919985e-07, "loss": 0.0008, "step": 22510 }, { "epoch": 6.99, "learning_rate": 9.993998799759954e-07, "loss": 0.0409, "step": 22511 }, { "epoch": 6.99, "learning_rate": 9.98999799959992e-07, "loss": 0.0018, "step": 22512 }, { "epoch": 6.99, "learning_rate": 9.985997199439888e-07, "loss": 0.0162, "step": 22513 }, { "epoch": 6.99, "learning_rate": 9.981996399279857e-07, "loss": 0.0014, "step": 22514 }, { "epoch": 6.99, "learning_rate": 9.977995599119824e-07, "loss": 0.0012, "step": 22515 }, { "epoch": 6.99, "learning_rate": 9.973994798959793e-07, "loss": 0.0024, "step": 22516 }, { "epoch": 6.99, "learning_rate": 9.969993998799762e-07, "loss": 0.0066, "step": 22517 }, { "epoch": 6.99, "learning_rate": 9.965993198639728e-07, "loss": 0.0021, "step": 22518 }, { "epoch": 6.99, "learning_rate": 9.961992398479697e-07, "loss": 0.0475, "step": 22519 }, { "epoch": 6.99, "learning_rate": 9.957991598319664e-07, "loss": 0.001, "step": 22520 }, { "epoch": 6.99, "learning_rate": 9.953990798159633e-07, "loss": 0.0022, "step": 22521 }, { "epoch": 6.99, "learning_rate": 9.9499899979996e-07, "loss": 0.0024, "step": 22522 }, { "epoch": 6.99, "learning_rate": 9.94598919783957e-07, "loss": 0.0193, "step": 22523 }, { "epoch": 6.99, "learning_rate": 9.941988397679536e-07, "loss": 0.0007, "step": 22524 }, { "epoch": 6.99, "learning_rate": 9.937987597519505e-07, "loss": 0.0016, "step": 22525 }, { "epoch": 6.99, "learning_rate": 9.933986797359472e-07, "loss": 0.003, "step": 22526 }, { "epoch": 6.99, "learning_rate": 9.92998599719944e-07, "loss": 0.0016, "step": 22527 }, { "epoch": 6.99, "learning_rate": 9.92598519703941e-07, "loss": 0.0023, "step": 22528 }, { "epoch": 6.99, "learning_rate": 9.921984396879377e-07, "loss": 0.0011, "step": 22529 }, { "epoch": 6.99, "learning_rate": 9.917983596719346e-07, "loss": 0.0026, "step": 22530 }, { "epoch": 6.99, "learning_rate": 9.913982796559313e-07, "loss": 0.0011, "step": 22531 }, { "epoch": 6.99, "learning_rate": 9.90998199639928e-07, "loss": 0.001, "step": 22532 }, { "epoch": 6.99, "learning_rate": 9.905981196239248e-07, "loss": 0.0012, "step": 22533 }, { "epoch": 6.99, "learning_rate": 9.901980396079217e-07, "loss": 0.001, "step": 22534 }, { "epoch": 6.99, "learning_rate": 9.897979595919184e-07, "loss": 0.0059, "step": 22535 }, { "epoch": 6.99, "learning_rate": 9.893978795759153e-07, "loss": 0.0063, "step": 22536 }, { "epoch": 6.99, "learning_rate": 9.88997799559912e-07, "loss": 0.0008, "step": 22537 }, { "epoch": 7.0, "learning_rate": 9.88597719543909e-07, "loss": 0.0008, "step": 22538 }, { "epoch": 7.0, "learning_rate": 9.881976395279056e-07, "loss": 0.0025, "step": 22539 }, { "epoch": 7.0, "learning_rate": 9.877975595119025e-07, "loss": 0.004, "step": 22540 }, { "epoch": 7.0, "learning_rate": 9.873974794958994e-07, "loss": 0.002, "step": 22541 }, { "epoch": 7.0, "learning_rate": 9.86997399479896e-07, "loss": 0.0024, "step": 22542 }, { "epoch": 7.0, "learning_rate": 9.865973194638928e-07, "loss": 0.0024, "step": 22543 }, { "epoch": 7.0, "learning_rate": 9.861972394478897e-07, "loss": 0.0085, "step": 22544 }, { "epoch": 7.0, "learning_rate": 9.857971594318863e-07, "loss": 0.0016, "step": 22545 }, { "epoch": 7.0, "learning_rate": 9.853970794158832e-07, "loss": 0.0011, "step": 22546 }, { "epoch": 7.0, "learning_rate": 9.849969993998801e-07, "loss": 0.0063, "step": 22547 }, { "epoch": 7.0, "learning_rate": 9.845969193838768e-07, "loss": 0.0013, "step": 22548 }, { "epoch": 7.0, "learning_rate": 9.841968393678737e-07, "loss": 0.0008, "step": 22549 }, { "epoch": 7.0, "learning_rate": 9.837967593518704e-07, "loss": 0.004, "step": 22550 }, { "epoch": 7.0, "learning_rate": 9.833966793358673e-07, "loss": 0.0011, "step": 22551 }, { "epoch": 7.0, "learning_rate": 9.829965993198642e-07, "loss": 0.0021, "step": 22552 }, { "epoch": 7.0, "learning_rate": 9.825965193038609e-07, "loss": 0.0014, "step": 22553 }, { "epoch": 7.0, "learning_rate": 9.821964392878576e-07, "loss": 0.0011, "step": 22554 }, { "epoch": 7.0, "learning_rate": 9.817963592718545e-07, "loss": 0.0006, "step": 22555 }, { "epoch": 7.0, "learning_rate": 9.813962792558512e-07, "loss": 0.0009, "step": 22556 }, { "epoch": 7.0, "learning_rate": 9.80996199239848e-07, "loss": 0.0004, "step": 22557 }, { "epoch": 7.0, "learning_rate": 9.80596119223845e-07, "loss": 0.0005, "step": 22558 }, { "epoch": 7.0, "learning_rate": 9.801960392078416e-07, "loss": 0.0007, "step": 22559 }, { "epoch": 7.0, "learning_rate": 9.797959591918385e-07, "loss": 0.0011, "step": 22560 }, { "epoch": 7.0, "learning_rate": 9.793958791758352e-07, "loss": 0.0264, "step": 22561 }, { "epoch": 7.0, "learning_rate": 9.78995799159832e-07, "loss": 0.0008, "step": 22562 }, { "epoch": 7.0, "learning_rate": 9.785957191438288e-07, "loss": 0.0009, "step": 22563 }, { "epoch": 7.0, "learning_rate": 9.781956391278257e-07, "loss": 0.0074, "step": 22564 }, { "epoch": 7.0, "learning_rate": 9.777955591118224e-07, "loss": 0.0014, "step": 22565 }, { "epoch": 7.0, "learning_rate": 9.773954790958193e-07, "loss": 0.0009, "step": 22566 }, { "epoch": 7.0, "learning_rate": 9.76995399079816e-07, "loss": 0.0011, "step": 22567 }, { "epoch": 7.0, "learning_rate": 9.765953190638129e-07, "loss": 0.0006, "step": 22568 }, { "epoch": 7.0, "learning_rate": 9.761952390478096e-07, "loss": 0.0047, "step": 22569 }, { "epoch": 7.01, "learning_rate": 9.757951590318065e-07, "loss": 0.0005, "step": 22570 }, { "epoch": 7.01, "learning_rate": 9.753950790158034e-07, "loss": 0.0006, "step": 22571 }, { "epoch": 7.01, "learning_rate": 9.749949989998e-07, "loss": 0.002, "step": 22572 }, { "epoch": 7.01, "learning_rate": 9.745949189837967e-07, "loss": 0.0027, "step": 22573 }, { "epoch": 7.01, "learning_rate": 9.741948389677936e-07, "loss": 0.0008, "step": 22574 }, { "epoch": 7.01, "learning_rate": 9.737947589517905e-07, "loss": 0.0024, "step": 22575 }, { "epoch": 7.01, "learning_rate": 9.733946789357872e-07, "loss": 0.0004, "step": 22576 }, { "epoch": 7.01, "learning_rate": 9.729945989197841e-07, "loss": 0.0294, "step": 22577 }, { "epoch": 7.01, "learning_rate": 9.725945189037808e-07, "loss": 0.0027, "step": 22578 }, { "epoch": 7.01, "learning_rate": 9.721944388877777e-07, "loss": 0.0006, "step": 22579 }, { "epoch": 7.01, "learning_rate": 9.717943588717744e-07, "loss": 0.0005, "step": 22580 }, { "epoch": 7.01, "learning_rate": 9.713942788557713e-07, "loss": 0.0011, "step": 22581 }, { "epoch": 7.01, "learning_rate": 9.709941988397682e-07, "loss": 0.001, "step": 22582 }, { "epoch": 7.01, "learning_rate": 9.705941188237649e-07, "loss": 0.0008, "step": 22583 }, { "epoch": 7.01, "learning_rate": 9.701940388077616e-07, "loss": 0.0007, "step": 22584 }, { "epoch": 7.01, "learning_rate": 9.697939587917585e-07, "loss": 0.002, "step": 22585 }, { "epoch": 7.01, "learning_rate": 9.693938787757551e-07, "loss": 0.0009, "step": 22586 }, { "epoch": 7.01, "learning_rate": 9.68993798759752e-07, "loss": 0.0009, "step": 22587 }, { "epoch": 7.01, "learning_rate": 9.68593718743749e-07, "loss": 0.0006, "step": 22588 }, { "epoch": 7.01, "learning_rate": 9.681936387277456e-07, "loss": 0.0009, "step": 22589 }, { "epoch": 7.01, "learning_rate": 9.677935587117425e-07, "loss": 0.0014, "step": 22590 }, { "epoch": 7.01, "learning_rate": 9.673934786957392e-07, "loss": 0.0006, "step": 22591 }, { "epoch": 7.01, "learning_rate": 9.669933986797359e-07, "loss": 0.0006, "step": 22592 }, { "epoch": 7.01, "learning_rate": 9.665933186637328e-07, "loss": 0.0009, "step": 22593 }, { "epoch": 7.01, "learning_rate": 9.661932386477297e-07, "loss": 0.0021, "step": 22594 }, { "epoch": 7.01, "learning_rate": 9.657931586317264e-07, "loss": 0.0011, "step": 22595 }, { "epoch": 7.01, "learning_rate": 9.653930786157233e-07, "loss": 0.0012, "step": 22596 }, { "epoch": 7.01, "learning_rate": 9.6499299859972e-07, "loss": 0.0013, "step": 22597 }, { "epoch": 7.01, "learning_rate": 9.645929185837169e-07, "loss": 0.0053, "step": 22598 }, { "epoch": 7.01, "learning_rate": 9.641928385677135e-07, "loss": 0.012, "step": 22599 }, { "epoch": 7.01, "learning_rate": 9.637927585517104e-07, "loss": 0.0017, "step": 22600 }, { "epoch": 7.01, "learning_rate": 9.633926785357073e-07, "loss": 0.0017, "step": 22601 }, { "epoch": 7.02, "learning_rate": 9.62992598519704e-07, "loss": 0.0017, "step": 22602 }, { "epoch": 7.02, "learning_rate": 9.625925185037007e-07, "loss": 0.0014, "step": 22603 }, { "epoch": 7.02, "learning_rate": 9.621924384876976e-07, "loss": 0.0008, "step": 22604 }, { "epoch": 7.02, "learning_rate": 9.617923584716945e-07, "loss": 0.0016, "step": 22605 }, { "epoch": 7.02, "learning_rate": 9.613922784556912e-07, "loss": 0.0008, "step": 22606 }, { "epoch": 7.02, "learning_rate": 9.60992198439688e-07, "loss": 0.0027, "step": 22607 }, { "epoch": 7.02, "learning_rate": 9.605921184236848e-07, "loss": 0.0007, "step": 22608 }, { "epoch": 7.02, "learning_rate": 9.601920384076815e-07, "loss": 0.0076, "step": 22609 }, { "epoch": 7.02, "learning_rate": 9.597919583916784e-07, "loss": 0.0004, "step": 22610 }, { "epoch": 7.02, "learning_rate": 9.593918783756753e-07, "loss": 0.0047, "step": 22611 }, { "epoch": 7.02, "learning_rate": 9.589917983596722e-07, "loss": 0.0012, "step": 22612 }, { "epoch": 7.02, "learning_rate": 9.585917183436688e-07, "loss": 0.0005, "step": 22613 }, { "epoch": 7.02, "learning_rate": 9.581916383276655e-07, "loss": 0.0007, "step": 22614 }, { "epoch": 7.02, "learning_rate": 9.577915583116624e-07, "loss": 0.0008, "step": 22615 }, { "epoch": 7.02, "learning_rate": 9.573914782956591e-07, "loss": 0.0009, "step": 22616 }, { "epoch": 7.02, "learning_rate": 9.56991398279656e-07, "loss": 0.001, "step": 22617 }, { "epoch": 7.02, "learning_rate": 9.56591318263653e-07, "loss": 0.0017, "step": 22618 }, { "epoch": 7.02, "learning_rate": 9.561912382476496e-07, "loss": 0.0007, "step": 22619 }, { "epoch": 7.02, "learning_rate": 9.557911582316465e-07, "loss": 0.0021, "step": 22620 }, { "epoch": 7.02, "learning_rate": 9.553910782156432e-07, "loss": 0.0007, "step": 22621 }, { "epoch": 7.02, "learning_rate": 9.5499099819964e-07, "loss": 0.0133, "step": 22622 }, { "epoch": 7.02, "learning_rate": 9.545909181836368e-07, "loss": 0.0007, "step": 22623 }, { "epoch": 7.02, "learning_rate": 9.541908381676337e-07, "loss": 0.0015, "step": 22624 }, { "epoch": 7.02, "learning_rate": 9.537907581516303e-07, "loss": 0.0021, "step": 22625 }, { "epoch": 7.02, "learning_rate": 9.533906781356272e-07, "loss": 0.0014, "step": 22626 }, { "epoch": 7.02, "learning_rate": 9.52990598119624e-07, "loss": 0.0009, "step": 22627 }, { "epoch": 7.02, "learning_rate": 9.525905181036207e-07, "loss": 0.0006, "step": 22628 }, { "epoch": 7.02, "learning_rate": 9.521904380876176e-07, "loss": 0.0045, "step": 22629 }, { "epoch": 7.02, "learning_rate": 9.517903580716144e-07, "loss": 0.0011, "step": 22630 }, { "epoch": 7.02, "learning_rate": 9.513902780556113e-07, "loss": 0.0067, "step": 22631 }, { "epoch": 7.02, "learning_rate": 9.50990198039608e-07, "loss": 0.0008, "step": 22632 }, { "epoch": 7.02, "learning_rate": 9.505901180236048e-07, "loss": 0.0011, "step": 22633 }, { "epoch": 7.03, "learning_rate": 9.501900380076017e-07, "loss": 0.0006, "step": 22634 }, { "epoch": 7.03, "learning_rate": 9.497899579915984e-07, "loss": 0.0009, "step": 22635 }, { "epoch": 7.03, "learning_rate": 9.493898779755952e-07, "loss": 0.0011, "step": 22636 }, { "epoch": 7.03, "learning_rate": 9.489897979595921e-07, "loss": 0.0009, "step": 22637 }, { "epoch": 7.03, "learning_rate": 9.485897179435888e-07, "loss": 0.0011, "step": 22638 }, { "epoch": 7.03, "learning_rate": 9.481896379275855e-07, "loss": 0.001, "step": 22639 }, { "epoch": 7.03, "learning_rate": 9.477895579115824e-07, "loss": 0.001, "step": 22640 }, { "epoch": 7.03, "learning_rate": 9.473894778955791e-07, "loss": 0.0025, "step": 22641 }, { "epoch": 7.03, "learning_rate": 9.46989397879576e-07, "loss": 0.0008, "step": 22642 }, { "epoch": 7.03, "learning_rate": 9.465893178635728e-07, "loss": 0.0008, "step": 22643 }, { "epoch": 7.03, "learning_rate": 9.461892378475695e-07, "loss": 0.001, "step": 22644 }, { "epoch": 7.03, "learning_rate": 9.457891578315664e-07, "loss": 0.0007, "step": 22645 }, { "epoch": 7.03, "learning_rate": 9.453890778155632e-07, "loss": 0.0011, "step": 22646 }, { "epoch": 7.03, "learning_rate": 9.4498899779956e-07, "loss": 0.0009, "step": 22647 }, { "epoch": 7.03, "learning_rate": 9.445889177835568e-07, "loss": 0.0007, "step": 22648 }, { "epoch": 7.03, "learning_rate": 9.441888377675536e-07, "loss": 0.0009, "step": 22649 }, { "epoch": 7.03, "learning_rate": 9.437887577515504e-07, "loss": 0.0008, "step": 22650 }, { "epoch": 7.03, "learning_rate": 9.433886777355472e-07, "loss": 0.0014, "step": 22651 }, { "epoch": 7.03, "learning_rate": 9.42988597719544e-07, "loss": 0.0007, "step": 22652 }, { "epoch": 7.03, "learning_rate": 9.425885177035408e-07, "loss": 0.0132, "step": 22653 }, { "epoch": 7.03, "learning_rate": 9.421884376875376e-07, "loss": 0.0009, "step": 22654 }, { "epoch": 7.03, "learning_rate": 9.417883576715343e-07, "loss": 0.001, "step": 22655 }, { "epoch": 7.03, "learning_rate": 9.413882776555312e-07, "loss": 0.0008, "step": 22656 }, { "epoch": 7.03, "learning_rate": 9.40988197639528e-07, "loss": 0.0018, "step": 22657 }, { "epoch": 7.03, "learning_rate": 9.405881176235247e-07, "loss": 0.0008, "step": 22658 }, { "epoch": 7.03, "learning_rate": 9.401880376075216e-07, "loss": 0.0008, "step": 22659 }, { "epoch": 7.03, "learning_rate": 9.397879575915184e-07, "loss": 0.0004, "step": 22660 }, { "epoch": 7.03, "learning_rate": 9.393878775755151e-07, "loss": 0.0006, "step": 22661 }, { "epoch": 7.03, "learning_rate": 9.38987797559512e-07, "loss": 0.0009, "step": 22662 }, { "epoch": 7.03, "learning_rate": 9.385877175435088e-07, "loss": 0.0185, "step": 22663 }, { "epoch": 7.03, "learning_rate": 9.381876375275057e-07, "loss": 0.0009, "step": 22664 }, { "epoch": 7.03, "learning_rate": 9.377875575115024e-07, "loss": 0.001, "step": 22665 }, { "epoch": 7.04, "learning_rate": 9.373874774954991e-07, "loss": 0.001, "step": 22666 }, { "epoch": 7.04, "learning_rate": 9.36987397479496e-07, "loss": 0.001, "step": 22667 }, { "epoch": 7.04, "learning_rate": 9.365873174634927e-07, "loss": 0.0252, "step": 22668 }, { "epoch": 7.04, "learning_rate": 9.361872374474895e-07, "loss": 0.0345, "step": 22669 }, { "epoch": 7.04, "learning_rate": 9.357871574314864e-07, "loss": 0.0007, "step": 22670 }, { "epoch": 7.04, "learning_rate": 9.353870774154831e-07, "loss": 0.0007, "step": 22671 }, { "epoch": 7.04, "learning_rate": 9.3498699739948e-07, "loss": 0.0016, "step": 22672 }, { "epoch": 7.04, "learning_rate": 9.345869173834768e-07, "loss": 0.0007, "step": 22673 }, { "epoch": 7.04, "learning_rate": 9.341868373674736e-07, "loss": 0.0037, "step": 22674 }, { "epoch": 7.04, "learning_rate": 9.337867573514704e-07, "loss": 0.0121, "step": 22675 }, { "epoch": 7.04, "learning_rate": 9.333866773354672e-07, "loss": 0.0151, "step": 22676 }, { "epoch": 7.04, "learning_rate": 9.32986597319464e-07, "loss": 0.0006, "step": 22677 }, { "epoch": 7.04, "learning_rate": 9.325865173034608e-07, "loss": 0.0009, "step": 22678 }, { "epoch": 7.04, "learning_rate": 9.321864372874575e-07, "loss": 0.0009, "step": 22679 }, { "epoch": 7.04, "learning_rate": 9.317863572714543e-07, "loss": 0.0011, "step": 22680 }, { "epoch": 7.04, "learning_rate": 9.313862772554512e-07, "loss": 0.0025, "step": 22681 }, { "epoch": 7.04, "learning_rate": 9.309861972394479e-07, "loss": 0.0007, "step": 22682 }, { "epoch": 7.04, "learning_rate": 9.305861172234448e-07, "loss": 0.0157, "step": 22683 }, { "epoch": 7.04, "learning_rate": 9.301860372074416e-07, "loss": 0.001, "step": 22684 }, { "epoch": 7.04, "learning_rate": 9.297859571914383e-07, "loss": 0.0007, "step": 22685 }, { "epoch": 7.04, "learning_rate": 9.293858771754352e-07, "loss": 0.0028, "step": 22686 }, { "epoch": 7.04, "learning_rate": 9.28985797159432e-07, "loss": 0.0014, "step": 22687 }, { "epoch": 7.04, "learning_rate": 9.285857171434287e-07, "loss": 0.0005, "step": 22688 }, { "epoch": 7.04, "learning_rate": 9.281856371274256e-07, "loss": 0.0009, "step": 22689 }, { "epoch": 7.04, "learning_rate": 9.277855571114224e-07, "loss": 0.0019, "step": 22690 }, { "epoch": 7.04, "learning_rate": 9.273854770954191e-07, "loss": 0.0011, "step": 22691 }, { "epoch": 7.04, "learning_rate": 9.26985397079416e-07, "loss": 0.0006, "step": 22692 }, { "epoch": 7.04, "learning_rate": 9.265853170634127e-07, "loss": 0.0008, "step": 22693 }, { "epoch": 7.04, "learning_rate": 9.261852370474096e-07, "loss": 0.0009, "step": 22694 }, { "epoch": 7.04, "learning_rate": 9.257851570314063e-07, "loss": 0.0008, "step": 22695 }, { "epoch": 7.04, "learning_rate": 9.253850770154031e-07, "loss": 0.0009, "step": 22696 }, { "epoch": 7.04, "learning_rate": 9.249849969994e-07, "loss": 0.0004, "step": 22697 }, { "epoch": 7.04, "learning_rate": 9.245849169833967e-07, "loss": 0.0005, "step": 22698 }, { "epoch": 7.05, "learning_rate": 9.241848369673935e-07, "loss": 0.0011, "step": 22699 }, { "epoch": 7.05, "learning_rate": 9.237847569513904e-07, "loss": 0.0007, "step": 22700 }, { "epoch": 7.05, "learning_rate": 9.233846769353872e-07, "loss": 0.0008, "step": 22701 }, { "epoch": 7.05, "learning_rate": 9.229845969193839e-07, "loss": 0.0011, "step": 22702 }, { "epoch": 7.05, "learning_rate": 9.225845169033808e-07, "loss": 0.0007, "step": 22703 }, { "epoch": 7.05, "learning_rate": 9.221844368873776e-07, "loss": 0.0012, "step": 22704 }, { "epoch": 7.05, "learning_rate": 9.217843568713744e-07, "loss": 0.0236, "step": 22705 }, { "epoch": 7.05, "learning_rate": 9.213842768553711e-07, "loss": 0.0012, "step": 22706 }, { "epoch": 7.05, "learning_rate": 9.209841968393679e-07, "loss": 0.0008, "step": 22707 }, { "epoch": 7.05, "learning_rate": 9.205841168233648e-07, "loss": 0.0237, "step": 22708 }, { "epoch": 7.05, "learning_rate": 9.201840368073615e-07, "loss": 0.0012, "step": 22709 }, { "epoch": 7.05, "learning_rate": 9.197839567913583e-07, "loss": 0.0008, "step": 22710 }, { "epoch": 7.05, "learning_rate": 9.193838767753552e-07, "loss": 0.0012, "step": 22711 }, { "epoch": 7.05, "learning_rate": 9.189837967593519e-07, "loss": 0.0006, "step": 22712 }, { "epoch": 7.05, "learning_rate": 9.185837167433488e-07, "loss": 0.0009, "step": 22713 }, { "epoch": 7.05, "learning_rate": 9.181836367273456e-07, "loss": 0.0007, "step": 22714 }, { "epoch": 7.05, "learning_rate": 9.177835567113423e-07, "loss": 0.0018, "step": 22715 }, { "epoch": 7.05, "learning_rate": 9.173834766953392e-07, "loss": 0.007, "step": 22716 }, { "epoch": 7.05, "learning_rate": 9.16983396679336e-07, "loss": 0.0252, "step": 22717 }, { "epoch": 7.05, "learning_rate": 9.165833166633327e-07, "loss": 0.0007, "step": 22718 }, { "epoch": 7.05, "learning_rate": 9.161832366473296e-07, "loss": 0.001, "step": 22719 }, { "epoch": 7.05, "learning_rate": 9.157831566313263e-07, "loss": 0.0194, "step": 22720 }, { "epoch": 7.05, "learning_rate": 9.15383076615323e-07, "loss": 0.001, "step": 22721 }, { "epoch": 7.05, "learning_rate": 9.149829965993199e-07, "loss": 0.0007, "step": 22722 }, { "epoch": 7.05, "learning_rate": 9.145829165833167e-07, "loss": 0.0012, "step": 22723 }, { "epoch": 7.05, "learning_rate": 9.141828365673136e-07, "loss": 0.0008, "step": 22724 }, { "epoch": 7.05, "learning_rate": 9.137827565513103e-07, "loss": 0.0007, "step": 22725 }, { "epoch": 7.05, "learning_rate": 9.133826765353071e-07, "loss": 0.0006, "step": 22726 }, { "epoch": 7.05, "learning_rate": 9.12982596519304e-07, "loss": 0.0009, "step": 22727 }, { "epoch": 7.05, "learning_rate": 9.125825165033008e-07, "loss": 0.0012, "step": 22728 }, { "epoch": 7.05, "learning_rate": 9.121824364872975e-07, "loss": 0.001, "step": 22729 }, { "epoch": 7.05, "learning_rate": 9.117823564712944e-07, "loss": 0.0006, "step": 22730 }, { "epoch": 7.06, "learning_rate": 9.113822764552912e-07, "loss": 0.001, "step": 22731 }, { "epoch": 7.06, "learning_rate": 9.109821964392879e-07, "loss": 0.0007, "step": 22732 }, { "epoch": 7.06, "learning_rate": 9.105821164232847e-07, "loss": 0.002, "step": 22733 }, { "epoch": 7.06, "learning_rate": 9.101820364072815e-07, "loss": 0.0007, "step": 22734 }, { "epoch": 7.06, "learning_rate": 9.097819563912784e-07, "loss": 0.0013, "step": 22735 }, { "epoch": 7.06, "learning_rate": 9.093818763752751e-07, "loss": 0.0008, "step": 22736 }, { "epoch": 7.06, "learning_rate": 9.089817963592719e-07, "loss": 0.001, "step": 22737 }, { "epoch": 7.06, "learning_rate": 9.085817163432688e-07, "loss": 0.0037, "step": 22738 }, { "epoch": 7.06, "learning_rate": 9.081816363272655e-07, "loss": 0.0007, "step": 22739 }, { "epoch": 7.06, "learning_rate": 9.077815563112623e-07, "loss": 0.001, "step": 22740 }, { "epoch": 7.06, "learning_rate": 9.073814762952592e-07, "loss": 0.0007, "step": 22741 }, { "epoch": 7.06, "learning_rate": 9.069813962792559e-07, "loss": 0.0009, "step": 22742 }, { "epoch": 7.06, "learning_rate": 9.065813162632527e-07, "loss": 0.0009, "step": 22743 }, { "epoch": 7.06, "learning_rate": 9.061812362472496e-07, "loss": 0.0006, "step": 22744 }, { "epoch": 7.06, "learning_rate": 9.057811562312463e-07, "loss": 0.0008, "step": 22745 }, { "epoch": 7.06, "learning_rate": 9.053810762152432e-07, "loss": 0.0104, "step": 22746 }, { "epoch": 7.06, "learning_rate": 9.049809961992399e-07, "loss": 0.0009, "step": 22747 }, { "epoch": 7.06, "learning_rate": 9.045809161832366e-07, "loss": 0.0017, "step": 22748 }, { "epoch": 7.06, "learning_rate": 9.041808361672335e-07, "loss": 0.0009, "step": 22749 }, { "epoch": 7.06, "learning_rate": 9.037807561512303e-07, "loss": 0.001, "step": 22750 }, { "epoch": 7.06, "learning_rate": 9.033806761352271e-07, "loss": 0.0006, "step": 22751 }, { "epoch": 7.06, "learning_rate": 9.029805961192239e-07, "loss": 0.0014, "step": 22752 }, { "epoch": 7.06, "learning_rate": 9.025805161032207e-07, "loss": 0.0007, "step": 22753 }, { "epoch": 7.06, "learning_rate": 9.021804360872175e-07, "loss": 0.0022, "step": 22754 }, { "epoch": 7.06, "learning_rate": 9.017803560712144e-07, "loss": 0.001, "step": 22755 }, { "epoch": 7.06, "learning_rate": 9.013802760552111e-07, "loss": 0.0007, "step": 22756 }, { "epoch": 7.06, "learning_rate": 9.00980196039208e-07, "loss": 0.0008, "step": 22757 }, { "epoch": 7.06, "learning_rate": 9.005801160232048e-07, "loss": 0.0011, "step": 22758 }, { "epoch": 7.06, "learning_rate": 9.001800360072015e-07, "loss": 0.0009, "step": 22759 }, { "epoch": 7.06, "learning_rate": 8.997799559911983e-07, "loss": 0.0015, "step": 22760 }, { "epoch": 7.06, "learning_rate": 8.993798759751951e-07, "loss": 0.0005, "step": 22761 }, { "epoch": 7.06, "learning_rate": 8.989797959591918e-07, "loss": 0.0008, "step": 22762 }, { "epoch": 7.07, "learning_rate": 8.985797159431887e-07, "loss": 0.001, "step": 22763 }, { "epoch": 7.07, "learning_rate": 8.981796359271855e-07, "loss": 0.0007, "step": 22764 }, { "epoch": 7.07, "learning_rate": 8.977795559111824e-07, "loss": 0.0006, "step": 22765 }, { "epoch": 7.07, "learning_rate": 8.973794758951791e-07, "loss": 0.0016, "step": 22766 }, { "epoch": 7.07, "learning_rate": 8.969793958791759e-07, "loss": 0.001, "step": 22767 }, { "epoch": 7.07, "learning_rate": 8.965793158631728e-07, "loss": 0.0004, "step": 22768 }, { "epoch": 7.07, "learning_rate": 8.961792358471695e-07, "loss": 0.0009, "step": 22769 }, { "epoch": 7.07, "learning_rate": 8.957791558311663e-07, "loss": 0.002, "step": 22770 }, { "epoch": 7.07, "learning_rate": 8.953790758151632e-07, "loss": 0.0005, "step": 22771 }, { "epoch": 7.07, "learning_rate": 8.949789957991599e-07, "loss": 0.0007, "step": 22772 }, { "epoch": 7.07, "learning_rate": 8.945789157831566e-07, "loss": 0.0006, "step": 22773 }, { "epoch": 7.07, "learning_rate": 8.941788357671535e-07, "loss": 0.001, "step": 22774 }, { "epoch": 7.07, "learning_rate": 8.937787557511502e-07, "loss": 0.0008, "step": 22775 }, { "epoch": 7.07, "learning_rate": 8.933786757351471e-07, "loss": 0.0007, "step": 22776 }, { "epoch": 7.07, "learning_rate": 8.929785957191439e-07, "loss": 0.0009, "step": 22777 }, { "epoch": 7.07, "learning_rate": 8.925785157031407e-07, "loss": 0.001, "step": 22778 }, { "epoch": 7.07, "learning_rate": 8.921784356871375e-07, "loss": 0.0005, "step": 22779 }, { "epoch": 7.07, "learning_rate": 8.917783556711343e-07, "loss": 0.0009, "step": 22780 }, { "epoch": 7.07, "learning_rate": 8.913782756551311e-07, "loss": 0.0013, "step": 22781 }, { "epoch": 7.07, "learning_rate": 8.90978195639128e-07, "loss": 0.0014, "step": 22782 }, { "epoch": 7.07, "learning_rate": 8.905781156231247e-07, "loss": 0.001, "step": 22783 }, { "epoch": 7.07, "learning_rate": 8.901780356071215e-07, "loss": 0.0005, "step": 22784 }, { "epoch": 7.07, "learning_rate": 8.897779555911184e-07, "loss": 0.0065, "step": 22785 }, { "epoch": 7.07, "learning_rate": 8.89377875575115e-07, "loss": 0.0007, "step": 22786 }, { "epoch": 7.07, "learning_rate": 8.88977795559112e-07, "loss": 0.0247, "step": 22787 }, { "epoch": 7.07, "learning_rate": 8.885777155431087e-07, "loss": 0.0009, "step": 22788 }, { "epoch": 7.07, "learning_rate": 8.881776355271054e-07, "loss": 0.0008, "step": 22789 }, { "epoch": 7.07, "learning_rate": 8.877775555111023e-07, "loss": 0.0013, "step": 22790 }, { "epoch": 7.07, "learning_rate": 8.873774754950991e-07, "loss": 0.0006, "step": 22791 }, { "epoch": 7.07, "learning_rate": 8.869773954790958e-07, "loss": 0.0011, "step": 22792 }, { "epoch": 7.07, "learning_rate": 8.865773154630927e-07, "loss": 0.0008, "step": 22793 }, { "epoch": 7.07, "learning_rate": 8.861772354470895e-07, "loss": 0.0003, "step": 22794 }, { "epoch": 7.08, "learning_rate": 8.857771554310862e-07, "loss": 0.001, "step": 22795 }, { "epoch": 7.08, "learning_rate": 8.853770754150831e-07, "loss": 0.0007, "step": 22796 }, { "epoch": 7.08, "learning_rate": 8.849769953990799e-07, "loss": 0.0008, "step": 22797 }, { "epoch": 7.08, "learning_rate": 8.845769153830768e-07, "loss": 0.0006, "step": 22798 }, { "epoch": 7.08, "learning_rate": 8.841768353670735e-07, "loss": 0.0005, "step": 22799 }, { "epoch": 7.08, "learning_rate": 8.837767553510702e-07, "loss": 0.0028, "step": 22800 }, { "epoch": 7.08, "learning_rate": 8.833766753350671e-07, "loss": 0.0011, "step": 22801 }, { "epoch": 7.08, "learning_rate": 8.829765953190638e-07, "loss": 0.0008, "step": 22802 }, { "epoch": 7.08, "learning_rate": 8.825765153030606e-07, "loss": 0.0007, "step": 22803 }, { "epoch": 7.08, "learning_rate": 8.821764352870575e-07, "loss": 0.001, "step": 22804 }, { "epoch": 7.08, "learning_rate": 8.817763552710543e-07, "loss": 0.0009, "step": 22805 }, { "epoch": 7.08, "learning_rate": 8.813762752550511e-07, "loss": 0.001, "step": 22806 }, { "epoch": 7.08, "learning_rate": 8.809761952390479e-07, "loss": 0.0006, "step": 22807 }, { "epoch": 7.08, "learning_rate": 8.805761152230447e-07, "loss": 0.0009, "step": 22808 }, { "epoch": 7.08, "learning_rate": 8.801760352070416e-07, "loss": 0.001, "step": 22809 }, { "epoch": 7.08, "learning_rate": 8.797759551910383e-07, "loss": 0.0005, "step": 22810 }, { "epoch": 7.08, "learning_rate": 8.793758751750351e-07, "loss": 0.0009, "step": 22811 }, { "epoch": 7.08, "learning_rate": 8.78975795159032e-07, "loss": 0.0013, "step": 22812 }, { "epoch": 7.08, "learning_rate": 8.785757151430287e-07, "loss": 0.0012, "step": 22813 }, { "epoch": 7.08, "learning_rate": 8.781756351270254e-07, "loss": 0.003, "step": 22814 }, { "epoch": 7.08, "learning_rate": 8.777755551110223e-07, "loss": 0.0011, "step": 22815 }, { "epoch": 7.08, "learning_rate": 8.77375475095019e-07, "loss": 0.0119, "step": 22816 }, { "epoch": 7.08, "learning_rate": 8.769753950790159e-07, "loss": 0.0011, "step": 22817 }, { "epoch": 7.08, "learning_rate": 8.765753150630127e-07, "loss": 0.0007, "step": 22818 }, { "epoch": 7.08, "learning_rate": 8.761752350470094e-07, "loss": 0.0008, "step": 22819 }, { "epoch": 7.08, "learning_rate": 8.757751550310063e-07, "loss": 0.0009, "step": 22820 }, { "epoch": 7.08, "learning_rate": 8.753750750150031e-07, "loss": 0.0008, "step": 22821 }, { "epoch": 7.08, "learning_rate": 8.749749949989998e-07, "loss": 0.0006, "step": 22822 }, { "epoch": 7.08, "learning_rate": 8.745749149829967e-07, "loss": 0.0012, "step": 22823 }, { "epoch": 7.08, "learning_rate": 8.741748349669935e-07, "loss": 0.0032, "step": 22824 }, { "epoch": 7.08, "learning_rate": 8.737747549509902e-07, "loss": 0.0009, "step": 22825 }, { "epoch": 7.08, "learning_rate": 8.733746749349871e-07, "loss": 0.0014, "step": 22826 }, { "epoch": 7.09, "learning_rate": 8.729745949189838e-07, "loss": 0.0008, "step": 22827 }, { "epoch": 7.09, "learning_rate": 8.725745149029807e-07, "loss": 0.0008, "step": 22828 }, { "epoch": 7.09, "learning_rate": 8.721744348869774e-07, "loss": 0.001, "step": 22829 }, { "epoch": 7.09, "learning_rate": 8.717743548709742e-07, "loss": 0.0017, "step": 22830 }, { "epoch": 7.09, "learning_rate": 8.713742748549711e-07, "loss": 0.0021, "step": 22831 }, { "epoch": 7.09, "learning_rate": 8.709741948389679e-07, "loss": 0.0082, "step": 22832 }, { "epoch": 7.09, "learning_rate": 8.705741148229646e-07, "loss": 0.0018, "step": 22833 }, { "epoch": 7.09, "learning_rate": 8.701740348069615e-07, "loss": 0.0021, "step": 22834 }, { "epoch": 7.09, "learning_rate": 8.697739547909583e-07, "loss": 0.001, "step": 22835 }, { "epoch": 7.09, "learning_rate": 8.69373874774955e-07, "loss": 0.0061, "step": 22836 }, { "epoch": 7.09, "learning_rate": 8.689737947589519e-07, "loss": 0.0008, "step": 22837 }, { "epoch": 7.09, "learning_rate": 8.685737147429487e-07, "loss": 0.0367, "step": 22838 }, { "epoch": 7.09, "learning_rate": 8.681736347269456e-07, "loss": 0.0006, "step": 22839 }, { "epoch": 7.09, "learning_rate": 8.677735547109423e-07, "loss": 0.0006, "step": 22840 }, { "epoch": 7.09, "learning_rate": 8.67373474694939e-07, "loss": 0.0007, "step": 22841 }, { "epoch": 7.09, "learning_rate": 8.669733946789359e-07, "loss": 0.001, "step": 22842 }, { "epoch": 7.09, "learning_rate": 8.665733146629326e-07, "loss": 0.0011, "step": 22843 }, { "epoch": 7.09, "learning_rate": 8.661732346469294e-07, "loss": 0.0021, "step": 22844 }, { "epoch": 7.09, "learning_rate": 8.657731546309263e-07, "loss": 0.0049, "step": 22845 }, { "epoch": 7.09, "learning_rate": 8.65373074614923e-07, "loss": 0.0013, "step": 22846 }, { "epoch": 7.09, "learning_rate": 8.649729945989199e-07, "loss": 0.0013, "step": 22847 }, { "epoch": 7.09, "learning_rate": 8.645729145829167e-07, "loss": 0.0016, "step": 22848 }, { "epoch": 7.09, "learning_rate": 8.641728345669134e-07, "loss": 0.001, "step": 22849 }, { "epoch": 7.09, "learning_rate": 8.637727545509103e-07, "loss": 0.0004, "step": 22850 }, { "epoch": 7.09, "learning_rate": 8.633726745349071e-07, "loss": 0.0005, "step": 22851 }, { "epoch": 7.09, "learning_rate": 8.629725945189038e-07, "loss": 0.0004, "step": 22852 }, { "epoch": 7.09, "learning_rate": 8.625725145029007e-07, "loss": 0.0025, "step": 22853 }, { "epoch": 7.09, "learning_rate": 8.621724344868974e-07, "loss": 0.0014, "step": 22854 }, { "epoch": 7.09, "learning_rate": 8.617723544708942e-07, "loss": 0.0021, "step": 22855 }, { "epoch": 7.09, "learning_rate": 8.61372274454891e-07, "loss": 0.0007, "step": 22856 }, { "epoch": 7.09, "learning_rate": 8.609721944388878e-07, "loss": 0.0441, "step": 22857 }, { "epoch": 7.09, "learning_rate": 8.605721144228847e-07, "loss": 0.0011, "step": 22858 }, { "epoch": 7.09, "learning_rate": 8.601720344068815e-07, "loss": 0.0006, "step": 22859 }, { "epoch": 7.1, "learning_rate": 8.597719543908782e-07, "loss": 0.0007, "step": 22860 }, { "epoch": 7.1, "learning_rate": 8.593718743748751e-07, "loss": 0.0008, "step": 22861 }, { "epoch": 7.1, "learning_rate": 8.589717943588719e-07, "loss": 0.0011, "step": 22862 }, { "epoch": 7.1, "learning_rate": 8.585717143428686e-07, "loss": 0.0008, "step": 22863 }, { "epoch": 7.1, "learning_rate": 8.581716343268655e-07, "loss": 0.0007, "step": 22864 }, { "epoch": 7.1, "learning_rate": 8.577715543108623e-07, "loss": 0.0015, "step": 22865 }, { "epoch": 7.1, "learning_rate": 8.57371474294859e-07, "loss": 0.0013, "step": 22866 }, { "epoch": 7.1, "learning_rate": 8.569713942788559e-07, "loss": 0.0014, "step": 22867 }, { "epoch": 7.1, "learning_rate": 8.565713142628526e-07, "loss": 0.0014, "step": 22868 }, { "epoch": 7.1, "learning_rate": 8.561712342468495e-07, "loss": 0.0016, "step": 22869 }, { "epoch": 7.1, "learning_rate": 8.557711542308462e-07, "loss": 0.0008, "step": 22870 }, { "epoch": 7.1, "learning_rate": 8.55371074214843e-07, "loss": 0.0008, "step": 22871 }, { "epoch": 7.1, "learning_rate": 8.549709941988399e-07, "loss": 0.0013, "step": 22872 }, { "epoch": 7.1, "learning_rate": 8.545709141828366e-07, "loss": 0.0275, "step": 22873 }, { "epoch": 7.1, "learning_rate": 8.541708341668334e-07, "loss": 0.0006, "step": 22874 }, { "epoch": 7.1, "learning_rate": 8.537707541508303e-07, "loss": 0.0014, "step": 22875 }, { "epoch": 7.1, "learning_rate": 8.53370674134827e-07, "loss": 0.0008, "step": 22876 }, { "epoch": 7.1, "learning_rate": 8.529705941188238e-07, "loss": 0.0009, "step": 22877 }, { "epoch": 7.1, "learning_rate": 8.525705141028207e-07, "loss": 0.0016, "step": 22878 }, { "epoch": 7.1, "learning_rate": 8.521704340868174e-07, "loss": 0.0004, "step": 22879 }, { "epoch": 7.1, "learning_rate": 8.517703540708143e-07, "loss": 0.0006, "step": 22880 }, { "epoch": 7.1, "learning_rate": 8.51370274054811e-07, "loss": 0.0024, "step": 22881 }, { "epoch": 7.1, "learning_rate": 8.509701940388078e-07, "loss": 0.0004, "step": 22882 }, { "epoch": 7.1, "learning_rate": 8.505701140228046e-07, "loss": 0.0009, "step": 22883 }, { "epoch": 7.1, "learning_rate": 8.501700340068014e-07, "loss": 0.0118, "step": 22884 }, { "epoch": 7.1, "learning_rate": 8.497699539907982e-07, "loss": 0.0009, "step": 22885 }, { "epoch": 7.1, "learning_rate": 8.493698739747951e-07, "loss": 0.0012, "step": 22886 }, { "epoch": 7.1, "learning_rate": 8.489697939587918e-07, "loss": 0.0015, "step": 22887 }, { "epoch": 7.1, "learning_rate": 8.485697139427886e-07, "loss": 0.0008, "step": 22888 }, { "epoch": 7.1, "learning_rate": 8.481696339267855e-07, "loss": 0.0174, "step": 22889 }, { "epoch": 7.1, "learning_rate": 8.477695539107822e-07, "loss": 0.0016, "step": 22890 }, { "epoch": 7.1, "learning_rate": 8.473694738947791e-07, "loss": 0.0009, "step": 22891 }, { "epoch": 7.11, "learning_rate": 8.469693938787759e-07, "loss": 0.0006, "step": 22892 }, { "epoch": 7.11, "learning_rate": 8.465693138627726e-07, "loss": 0.0005, "step": 22893 }, { "epoch": 7.11, "learning_rate": 8.461692338467695e-07, "loss": 0.0008, "step": 22894 }, { "epoch": 7.11, "learning_rate": 8.457691538307662e-07, "loss": 0.0008, "step": 22895 }, { "epoch": 7.11, "learning_rate": 8.453690738147629e-07, "loss": 0.0011, "step": 22896 }, { "epoch": 7.11, "learning_rate": 8.449689937987598e-07, "loss": 0.0011, "step": 22897 }, { "epoch": 7.11, "learning_rate": 8.445689137827566e-07, "loss": 0.0018, "step": 22898 }, { "epoch": 7.11, "learning_rate": 8.441688337667535e-07, "loss": 0.0005, "step": 22899 }, { "epoch": 7.11, "learning_rate": 8.437687537507502e-07, "loss": 0.0005, "step": 22900 }, { "epoch": 7.11, "learning_rate": 8.43368673734747e-07, "loss": 0.0013, "step": 22901 }, { "epoch": 7.11, "learning_rate": 8.429685937187439e-07, "loss": 0.0121, "step": 22902 }, { "epoch": 7.11, "learning_rate": 8.425685137027406e-07, "loss": 0.0005, "step": 22903 }, { "epoch": 7.11, "learning_rate": 8.421684336867374e-07, "loss": 0.0011, "step": 22904 }, { "epoch": 7.11, "learning_rate": 8.417683536707343e-07, "loss": 0.0006, "step": 22905 }, { "epoch": 7.11, "learning_rate": 8.41368273654731e-07, "loss": 0.0007, "step": 22906 }, { "epoch": 7.11, "learning_rate": 8.409681936387277e-07, "loss": 0.0019, "step": 22907 }, { "epoch": 7.11, "learning_rate": 8.405681136227246e-07, "loss": 0.0011, "step": 22908 }, { "epoch": 7.11, "learning_rate": 8.401680336067214e-07, "loss": 0.0007, "step": 22909 }, { "epoch": 7.11, "learning_rate": 8.397679535907182e-07, "loss": 0.0006, "step": 22910 }, { "epoch": 7.11, "learning_rate": 8.39367873574715e-07, "loss": 0.0005, "step": 22911 }, { "epoch": 7.11, "learning_rate": 8.389677935587118e-07, "loss": 0.0007, "step": 22912 }, { "epoch": 7.11, "learning_rate": 8.385677135427087e-07, "loss": 0.0008, "step": 22913 }, { "epoch": 7.11, "learning_rate": 8.381676335267054e-07, "loss": 0.0004, "step": 22914 }, { "epoch": 7.11, "learning_rate": 8.377675535107022e-07, "loss": 0.0014, "step": 22915 }, { "epoch": 7.11, "learning_rate": 8.373674734946991e-07, "loss": 0.0012, "step": 22916 }, { "epoch": 7.11, "learning_rate": 8.369673934786958e-07, "loss": 0.0009, "step": 22917 }, { "epoch": 7.11, "learning_rate": 8.365673134626926e-07, "loss": 0.0013, "step": 22918 }, { "epoch": 7.11, "learning_rate": 8.361672334466895e-07, "loss": 0.0025, "step": 22919 }, { "epoch": 7.11, "learning_rate": 8.357671534306862e-07, "loss": 0.0008, "step": 22920 }, { "epoch": 7.11, "learning_rate": 8.35367073414683e-07, "loss": 0.0006, "step": 22921 }, { "epoch": 7.11, "learning_rate": 8.349669933986798e-07, "loss": 0.0009, "step": 22922 }, { "epoch": 7.11, "learning_rate": 8.345669133826765e-07, "loss": 0.0019, "step": 22923 }, { "epoch": 7.12, "learning_rate": 8.341668333666734e-07, "loss": 0.0005, "step": 22924 }, { "epoch": 7.12, "learning_rate": 8.337667533506702e-07, "loss": 0.0012, "step": 22925 }, { "epoch": 7.12, "learning_rate": 8.333666733346669e-07, "loss": 0.0009, "step": 22926 }, { "epoch": 7.12, "learning_rate": 8.329665933186638e-07, "loss": 0.0129, "step": 22927 }, { "epoch": 7.12, "learning_rate": 8.325665133026606e-07, "loss": 0.0011, "step": 22928 }, { "epoch": 7.12, "learning_rate": 8.321664332866573e-07, "loss": 0.0016, "step": 22929 }, { "epoch": 7.12, "learning_rate": 8.317663532706542e-07, "loss": 0.0012, "step": 22930 }, { "epoch": 7.12, "learning_rate": 8.31366273254651e-07, "loss": 0.0216, "step": 22931 }, { "epoch": 7.12, "learning_rate": 8.309661932386479e-07, "loss": 0.0009, "step": 22932 }, { "epoch": 7.12, "learning_rate": 8.305661132226446e-07, "loss": 0.0009, "step": 22933 }, { "epoch": 7.12, "learning_rate": 8.301660332066413e-07, "loss": 0.0007, "step": 22934 }, { "epoch": 7.12, "learning_rate": 8.297659531906382e-07, "loss": 0.0004, "step": 22935 }, { "epoch": 7.12, "learning_rate": 8.29365873174635e-07, "loss": 0.0005, "step": 22936 }, { "epoch": 7.12, "learning_rate": 8.289657931586317e-07, "loss": 0.0004, "step": 22937 }, { "epoch": 7.12, "learning_rate": 8.285657131426286e-07, "loss": 0.0012, "step": 22938 }, { "epoch": 7.12, "learning_rate": 8.281656331266254e-07, "loss": 0.0015, "step": 22939 }, { "epoch": 7.12, "learning_rate": 8.277655531106223e-07, "loss": 0.0005, "step": 22940 }, { "epoch": 7.12, "learning_rate": 8.27365473094619e-07, "loss": 0.0019, "step": 22941 }, { "epoch": 7.12, "learning_rate": 8.269653930786158e-07, "loss": 0.001, "step": 22942 }, { "epoch": 7.12, "learning_rate": 8.265653130626127e-07, "loss": 0.001, "step": 22943 }, { "epoch": 7.12, "learning_rate": 8.261652330466094e-07, "loss": 0.0008, "step": 22944 }, { "epoch": 7.12, "learning_rate": 8.257651530306062e-07, "loss": 0.001, "step": 22945 }, { "epoch": 7.12, "learning_rate": 8.253650730146031e-07, "loss": 0.0011, "step": 22946 }, { "epoch": 7.12, "learning_rate": 8.249649929985998e-07, "loss": 0.0008, "step": 22947 }, { "epoch": 7.12, "learning_rate": 8.245649129825965e-07, "loss": 0.0005, "step": 22948 }, { "epoch": 7.12, "learning_rate": 8.241648329665934e-07, "loss": 0.0007, "step": 22949 }, { "epoch": 7.12, "learning_rate": 8.237647529505901e-07, "loss": 0.0009, "step": 22950 }, { "epoch": 7.12, "learning_rate": 8.23364672934587e-07, "loss": 0.0006, "step": 22951 }, { "epoch": 7.12, "learning_rate": 8.229645929185838e-07, "loss": 0.0011, "step": 22952 }, { "epoch": 7.12, "learning_rate": 8.225645129025805e-07, "loss": 0.0005, "step": 22953 }, { "epoch": 7.12, "learning_rate": 8.221644328865774e-07, "loss": 0.0004, "step": 22954 }, { "epoch": 7.12, "learning_rate": 8.217643528705742e-07, "loss": 0.0011, "step": 22955 }, { "epoch": 7.13, "learning_rate": 8.213642728545709e-07, "loss": 0.0012, "step": 22956 }, { "epoch": 7.13, "learning_rate": 8.209641928385678e-07, "loss": 0.0004, "step": 22957 }, { "epoch": 7.13, "learning_rate": 8.205641128225646e-07, "loss": 0.0009, "step": 22958 }, { "epoch": 7.13, "learning_rate": 8.201640328065614e-07, "loss": 0.0013, "step": 22959 }, { "epoch": 7.13, "learning_rate": 8.197639527905582e-07, "loss": 0.0008, "step": 22960 }, { "epoch": 7.13, "learning_rate": 8.19363872774555e-07, "loss": 0.0008, "step": 22961 }, { "epoch": 7.13, "learning_rate": 8.189637927585518e-07, "loss": 0.0009, "step": 22962 }, { "epoch": 7.13, "learning_rate": 8.185637127425486e-07, "loss": 0.0009, "step": 22963 }, { "epoch": 7.13, "learning_rate": 8.181636327265453e-07, "loss": 0.001, "step": 22964 }, { "epoch": 7.13, "learning_rate": 8.177635527105422e-07, "loss": 0.0031, "step": 22965 }, { "epoch": 7.13, "learning_rate": 8.17363472694539e-07, "loss": 0.0307, "step": 22966 }, { "epoch": 7.13, "learning_rate": 8.169633926785357e-07, "loss": 0.0012, "step": 22967 }, { "epoch": 7.13, "learning_rate": 8.165633126625326e-07, "loss": 0.0005, "step": 22968 }, { "epoch": 7.13, "learning_rate": 8.161632326465294e-07, "loss": 0.0017, "step": 22969 }, { "epoch": 7.13, "learning_rate": 8.157631526305261e-07, "loss": 0.0091, "step": 22970 }, { "epoch": 7.13, "learning_rate": 8.15363072614523e-07, "loss": 0.001, "step": 22971 }, { "epoch": 7.13, "learning_rate": 8.149629925985198e-07, "loss": 0.0008, "step": 22972 }, { "epoch": 7.13, "learning_rate": 8.145629125825167e-07, "loss": 0.0006, "step": 22973 }, { "epoch": 7.13, "learning_rate": 8.141628325665134e-07, "loss": 0.0056, "step": 22974 }, { "epoch": 7.13, "learning_rate": 8.137627525505101e-07, "loss": 0.0011, "step": 22975 }, { "epoch": 7.13, "learning_rate": 8.13362672534507e-07, "loss": 0.0007, "step": 22976 }, { "epoch": 7.13, "learning_rate": 8.129625925185037e-07, "loss": 0.0013, "step": 22977 }, { "epoch": 7.13, "learning_rate": 8.125625125025005e-07, "loss": 0.0008, "step": 22978 }, { "epoch": 7.13, "learning_rate": 8.121624324864974e-07, "loss": 0.0013, "step": 22979 }, { "epoch": 7.13, "learning_rate": 8.117623524704941e-07, "loss": 0.0015, "step": 22980 }, { "epoch": 7.13, "learning_rate": 8.113622724544909e-07, "loss": 0.0011, "step": 22981 }, { "epoch": 7.13, "learning_rate": 8.109621924384878e-07, "loss": 0.0164, "step": 22982 }, { "epoch": 7.13, "learning_rate": 8.105621124224845e-07, "loss": 0.0063, "step": 22983 }, { "epoch": 7.13, "learning_rate": 8.101620324064814e-07, "loss": 0.001, "step": 22984 }, { "epoch": 7.13, "learning_rate": 8.097619523904782e-07, "loss": 0.0155, "step": 22985 }, { "epoch": 7.13, "learning_rate": 8.09361872374475e-07, "loss": 0.0613, "step": 22986 }, { "epoch": 7.13, "learning_rate": 8.089617923584718e-07, "loss": 0.0007, "step": 22987 }, { "epoch": 7.13, "learning_rate": 8.085617123424685e-07, "loss": 0.0011, "step": 22988 }, { "epoch": 7.14, "learning_rate": 8.081616323264653e-07, "loss": 0.001, "step": 22989 }, { "epoch": 7.14, "learning_rate": 8.077615523104622e-07, "loss": 0.0006, "step": 22990 }, { "epoch": 7.14, "learning_rate": 8.073614722944589e-07, "loss": 0.0006, "step": 22991 }, { "epoch": 7.14, "learning_rate": 8.069613922784558e-07, "loss": 0.0067, "step": 22992 }, { "epoch": 7.14, "learning_rate": 8.065613122624526e-07, "loss": 0.0009, "step": 22993 }, { "epoch": 7.14, "learning_rate": 8.061612322464493e-07, "loss": 0.0024, "step": 22994 }, { "epoch": 7.14, "learning_rate": 8.057611522304462e-07, "loss": 0.0005, "step": 22995 }, { "epoch": 7.14, "learning_rate": 8.05361072214443e-07, "loss": 0.0004, "step": 22996 }, { "epoch": 7.14, "learning_rate": 8.049609921984397e-07, "loss": 0.0006, "step": 22997 }, { "epoch": 7.14, "learning_rate": 8.045609121824366e-07, "loss": 0.0009, "step": 22998 }, { "epoch": 7.14, "learning_rate": 8.041608321664334e-07, "loss": 0.0013, "step": 22999 }, { "epoch": 7.14, "learning_rate": 8.037607521504301e-07, "loss": 0.0021, "step": 23000 }, { "epoch": 7.14, "learning_rate": 8.03360672134427e-07, "loss": 0.003, "step": 23001 }, { "epoch": 7.14, "learning_rate": 8.029605921184237e-07, "loss": 0.0011, "step": 23002 }, { "epoch": 7.14, "learning_rate": 8.025605121024206e-07, "loss": 0.001, "step": 23003 }, { "epoch": 7.14, "learning_rate": 8.021604320864173e-07, "loss": 0.0007, "step": 23004 }, { "epoch": 7.14, "learning_rate": 8.017603520704141e-07, "loss": 0.0014, "step": 23005 }, { "epoch": 7.14, "learning_rate": 8.01360272054411e-07, "loss": 0.0011, "step": 23006 }, { "epoch": 7.14, "learning_rate": 8.009601920384077e-07, "loss": 0.0009, "step": 23007 }, { "epoch": 7.14, "learning_rate": 8.005601120224045e-07, "loss": 0.0017, "step": 23008 }, { "epoch": 7.14, "learning_rate": 8.001600320064014e-07, "loss": 0.0014, "step": 23009 }, { "epoch": 7.14, "learning_rate": 7.997599519903981e-07, "loss": 0.0008, "step": 23010 }, { "epoch": 7.14, "learning_rate": 7.993598719743949e-07, "loss": 0.0014, "step": 23011 }, { "epoch": 7.14, "learning_rate": 7.989597919583918e-07, "loss": 0.0009, "step": 23012 }, { "epoch": 7.14, "learning_rate": 7.985597119423886e-07, "loss": 0.0009, "step": 23013 }, { "epoch": 7.14, "learning_rate": 7.981596319263854e-07, "loss": 0.0008, "step": 23014 }, { "epoch": 7.14, "learning_rate": 7.977595519103821e-07, "loss": 0.0006, "step": 23015 }, { "epoch": 7.14, "learning_rate": 7.973594718943789e-07, "loss": 0.0008, "step": 23016 }, { "epoch": 7.14, "learning_rate": 7.969593918783758e-07, "loss": 0.0009, "step": 23017 }, { "epoch": 7.14, "learning_rate": 7.965593118623725e-07, "loss": 0.0007, "step": 23018 }, { "epoch": 7.14, "learning_rate": 7.961592318463693e-07, "loss": 0.0012, "step": 23019 }, { "epoch": 7.14, "learning_rate": 7.957591518303662e-07, "loss": 0.001, "step": 23020 }, { "epoch": 7.15, "learning_rate": 7.953590718143629e-07, "loss": 0.0005, "step": 23021 }, { "epoch": 7.15, "learning_rate": 7.949589917983597e-07, "loss": 0.0006, "step": 23022 }, { "epoch": 7.15, "learning_rate": 7.945589117823566e-07, "loss": 0.0009, "step": 23023 }, { "epoch": 7.15, "learning_rate": 7.941588317663533e-07, "loss": 0.0009, "step": 23024 }, { "epoch": 7.15, "learning_rate": 7.937587517503502e-07, "loss": 0.0007, "step": 23025 }, { "epoch": 7.15, "learning_rate": 7.93358671734347e-07, "loss": 0.0008, "step": 23026 }, { "epoch": 7.15, "learning_rate": 7.929585917183437e-07, "loss": 0.0031, "step": 23027 }, { "epoch": 7.15, "learning_rate": 7.925585117023406e-07, "loss": 0.0068, "step": 23028 }, { "epoch": 7.15, "learning_rate": 7.921584316863373e-07, "loss": 0.0007, "step": 23029 }, { "epoch": 7.15, "learning_rate": 7.91758351670334e-07, "loss": 0.0018, "step": 23030 }, { "epoch": 7.15, "learning_rate": 7.913582716543309e-07, "loss": 0.0055, "step": 23031 }, { "epoch": 7.15, "learning_rate": 7.909581916383277e-07, "loss": 0.002, "step": 23032 }, { "epoch": 7.15, "learning_rate": 7.905581116223246e-07, "loss": 0.0007, "step": 23033 }, { "epoch": 7.15, "learning_rate": 7.901580316063213e-07, "loss": 0.0007, "step": 23034 }, { "epoch": 7.15, "learning_rate": 7.897579515903181e-07, "loss": 0.0008, "step": 23035 }, { "epoch": 7.15, "learning_rate": 7.89357871574315e-07, "loss": 0.0008, "step": 23036 }, { "epoch": 7.15, "learning_rate": 7.889577915583117e-07, "loss": 0.0006, "step": 23037 }, { "epoch": 7.15, "learning_rate": 7.885577115423085e-07, "loss": 0.0008, "step": 23038 }, { "epoch": 7.15, "learning_rate": 7.881576315263054e-07, "loss": 0.0188, "step": 23039 }, { "epoch": 7.15, "learning_rate": 7.877575515103022e-07, "loss": 0.0008, "step": 23040 }, { "epoch": 7.15, "learning_rate": 7.873574714942989e-07, "loss": 0.0004, "step": 23041 }, { "epoch": 7.15, "learning_rate": 7.869573914782957e-07, "loss": 0.0015, "step": 23042 }, { "epoch": 7.15, "learning_rate": 7.865573114622925e-07, "loss": 0.0009, "step": 23043 }, { "epoch": 7.15, "learning_rate": 7.861572314462894e-07, "loss": 0.0011, "step": 23044 }, { "epoch": 7.15, "learning_rate": 7.857571514302861e-07, "loss": 0.0007, "step": 23045 }, { "epoch": 7.15, "learning_rate": 7.853570714142829e-07, "loss": 0.0006, "step": 23046 }, { "epoch": 7.15, "learning_rate": 7.849569913982798e-07, "loss": 0.0008, "step": 23047 }, { "epoch": 7.15, "learning_rate": 7.845569113822765e-07, "loss": 0.0007, "step": 23048 }, { "epoch": 7.15, "learning_rate": 7.841568313662733e-07, "loss": 0.0007, "step": 23049 }, { "epoch": 7.15, "learning_rate": 7.837567513502702e-07, "loss": 0.0008, "step": 23050 }, { "epoch": 7.15, "learning_rate": 7.833566713342669e-07, "loss": 0.0006, "step": 23051 }, { "epoch": 7.15, "learning_rate": 7.829565913182637e-07, "loss": 0.0014, "step": 23052 }, { "epoch": 7.16, "learning_rate": 7.825565113022606e-07, "loss": 0.0003, "step": 23053 }, { "epoch": 7.16, "learning_rate": 7.821564312862573e-07, "loss": 0.001, "step": 23054 }, { "epoch": 7.16, "learning_rate": 7.817563512702542e-07, "loss": 0.0006, "step": 23055 }, { "epoch": 7.16, "learning_rate": 7.813562712542509e-07, "loss": 0.0007, "step": 23056 }, { "epoch": 7.16, "learning_rate": 7.809561912382476e-07, "loss": 0.0007, "step": 23057 }, { "epoch": 7.16, "learning_rate": 7.805561112222445e-07, "loss": 0.0011, "step": 23058 }, { "epoch": 7.16, "learning_rate": 7.801560312062413e-07, "loss": 0.0007, "step": 23059 }, { "epoch": 7.16, "learning_rate": 7.79755951190238e-07, "loss": 0.0011, "step": 23060 }, { "epoch": 7.16, "learning_rate": 7.793558711742349e-07, "loss": 0.0068, "step": 23061 }, { "epoch": 7.16, "learning_rate": 7.789557911582317e-07, "loss": 0.0004, "step": 23062 }, { "epoch": 7.16, "learning_rate": 7.785557111422285e-07, "loss": 0.0005, "step": 23063 }, { "epoch": 7.16, "learning_rate": 7.781556311262253e-07, "loss": 0.0006, "step": 23064 }, { "epoch": 7.16, "learning_rate": 7.777555511102221e-07, "loss": 0.0007, "step": 23065 }, { "epoch": 7.16, "learning_rate": 7.77355471094219e-07, "loss": 0.0009, "step": 23066 }, { "epoch": 7.16, "learning_rate": 7.769553910782158e-07, "loss": 0.0005, "step": 23067 }, { "epoch": 7.16, "learning_rate": 7.765553110622124e-07, "loss": 0.0073, "step": 23068 }, { "epoch": 7.16, "learning_rate": 7.761552310462093e-07, "loss": 0.0044, "step": 23069 }, { "epoch": 7.16, "learning_rate": 7.757551510302061e-07, "loss": 0.0028, "step": 23070 }, { "epoch": 7.16, "learning_rate": 7.753550710142028e-07, "loss": 0.0004, "step": 23071 }, { "epoch": 7.16, "learning_rate": 7.749549909981997e-07, "loss": 0.0008, "step": 23072 }, { "epoch": 7.16, "learning_rate": 7.745549109821965e-07, "loss": 0.001, "step": 23073 }, { "epoch": 7.16, "learning_rate": 7.741548309661932e-07, "loss": 0.0006, "step": 23074 }, { "epoch": 7.16, "learning_rate": 7.737547509501901e-07, "loss": 0.0009, "step": 23075 }, { "epoch": 7.16, "learning_rate": 7.733546709341869e-07, "loss": 0.0011, "step": 23076 }, { "epoch": 7.16, "learning_rate": 7.729545909181838e-07, "loss": 0.0008, "step": 23077 }, { "epoch": 7.16, "learning_rate": 7.725545109021805e-07, "loss": 0.0009, "step": 23078 }, { "epoch": 7.16, "learning_rate": 7.721544308861773e-07, "loss": 0.0009, "step": 23079 }, { "epoch": 7.16, "learning_rate": 7.717543508701742e-07, "loss": 0.0009, "step": 23080 }, { "epoch": 7.16, "learning_rate": 7.713542708541709e-07, "loss": 0.0012, "step": 23081 }, { "epoch": 7.16, "learning_rate": 7.709541908381676e-07, "loss": 0.0009, "step": 23082 }, { "epoch": 7.16, "learning_rate": 7.705541108221645e-07, "loss": 0.0008, "step": 23083 }, { "epoch": 7.16, "learning_rate": 7.701540308061612e-07, "loss": 0.0007, "step": 23084 }, { "epoch": 7.17, "learning_rate": 7.697539507901581e-07, "loss": 0.0006, "step": 23085 }, { "epoch": 7.17, "learning_rate": 7.693538707741549e-07, "loss": 0.0007, "step": 23086 }, { "epoch": 7.17, "learning_rate": 7.689537907581516e-07, "loss": 0.0008, "step": 23087 }, { "epoch": 7.17, "learning_rate": 7.685537107421485e-07, "loss": 0.0005, "step": 23088 }, { "epoch": 7.17, "learning_rate": 7.681536307261453e-07, "loss": 0.001, "step": 23089 }, { "epoch": 7.17, "learning_rate": 7.677535507101421e-07, "loss": 0.0009, "step": 23090 }, { "epoch": 7.17, "learning_rate": 7.673534706941389e-07, "loss": 0.001, "step": 23091 }, { "epoch": 7.17, "learning_rate": 7.669533906781357e-07, "loss": 0.0004, "step": 23092 }, { "epoch": 7.17, "learning_rate": 7.665533106621325e-07, "loss": 0.0004, "step": 23093 }, { "epoch": 7.17, "learning_rate": 7.661532306461294e-07, "loss": 0.0006, "step": 23094 }, { "epoch": 7.17, "learning_rate": 7.65753150630126e-07, "loss": 0.0008, "step": 23095 }, { "epoch": 7.17, "learning_rate": 7.65353070614123e-07, "loss": 0.0007, "step": 23096 }, { "epoch": 7.17, "learning_rate": 7.649529905981197e-07, "loss": 0.0006, "step": 23097 }, { "epoch": 7.17, "learning_rate": 7.645529105821164e-07, "loss": 0.0006, "step": 23098 }, { "epoch": 7.17, "learning_rate": 7.641528305661133e-07, "loss": 0.0014, "step": 23099 }, { "epoch": 7.17, "learning_rate": 7.637527505501101e-07, "loss": 0.011, "step": 23100 }, { "epoch": 7.17, "learning_rate": 7.633526705341068e-07, "loss": 0.0015, "step": 23101 }, { "epoch": 7.17, "learning_rate": 7.629525905181037e-07, "loss": 0.0155, "step": 23102 }, { "epoch": 7.17, "learning_rate": 7.625525105021005e-07, "loss": 0.0005, "step": 23103 }, { "epoch": 7.17, "learning_rate": 7.621524304860972e-07, "loss": 0.0144, "step": 23104 }, { "epoch": 7.17, "learning_rate": 7.617523504700941e-07, "loss": 0.0013, "step": 23105 }, { "epoch": 7.17, "learning_rate": 7.613522704540909e-07, "loss": 0.0008, "step": 23106 }, { "epoch": 7.17, "learning_rate": 7.609521904380878e-07, "loss": 0.0012, "step": 23107 }, { "epoch": 7.17, "learning_rate": 7.605521104220845e-07, "loss": 0.0009, "step": 23108 }, { "epoch": 7.17, "learning_rate": 7.601520304060812e-07, "loss": 0.029, "step": 23109 }, { "epoch": 7.17, "learning_rate": 7.597519503900781e-07, "loss": 0.002, "step": 23110 }, { "epoch": 7.17, "learning_rate": 7.593518703740748e-07, "loss": 0.0004, "step": 23111 }, { "epoch": 7.17, "learning_rate": 7.589517903580716e-07, "loss": 0.0011, "step": 23112 }, { "epoch": 7.17, "learning_rate": 7.585517103420685e-07, "loss": 0.0011, "step": 23113 }, { "epoch": 7.17, "learning_rate": 7.581516303260652e-07, "loss": 0.0007, "step": 23114 }, { "epoch": 7.17, "learning_rate": 7.57751550310062e-07, "loss": 0.0017, "step": 23115 }, { "epoch": 7.17, "learning_rate": 7.573514702940589e-07, "loss": 0.001, "step": 23116 }, { "epoch": 7.18, "learning_rate": 7.569513902780557e-07, "loss": 0.0006, "step": 23117 }, { "epoch": 7.18, "learning_rate": 7.565513102620525e-07, "loss": 0.0131, "step": 23118 }, { "epoch": 7.18, "learning_rate": 7.561512302460493e-07, "loss": 0.0007, "step": 23119 }, { "epoch": 7.18, "learning_rate": 7.557511502300461e-07, "loss": 0.0019, "step": 23120 }, { "epoch": 7.18, "learning_rate": 7.55351070214043e-07, "loss": 0.0006, "step": 23121 }, { "epoch": 7.18, "learning_rate": 7.549509901980396e-07, "loss": 0.0014, "step": 23122 }, { "epoch": 7.18, "learning_rate": 7.545509101820364e-07, "loss": 0.0008, "step": 23123 }, { "epoch": 7.18, "learning_rate": 7.541508301660333e-07, "loss": 0.0011, "step": 23124 }, { "epoch": 7.18, "learning_rate": 7.5375075015003e-07, "loss": 0.0012, "step": 23125 }, { "epoch": 7.18, "learning_rate": 7.533506701340269e-07, "loss": 0.0022, "step": 23126 }, { "epoch": 7.18, "learning_rate": 7.529505901180237e-07, "loss": 0.0009, "step": 23127 }, { "epoch": 7.18, "learning_rate": 7.525505101020204e-07, "loss": 0.0015, "step": 23128 }, { "epoch": 7.18, "learning_rate": 7.521504300860173e-07, "loss": 0.001, "step": 23129 }, { "epoch": 7.18, "learning_rate": 7.517503500700141e-07, "loss": 0.0047, "step": 23130 }, { "epoch": 7.18, "learning_rate": 7.513502700540108e-07, "loss": 0.0008, "step": 23131 }, { "epoch": 7.18, "learning_rate": 7.509501900380077e-07, "loss": 0.0008, "step": 23132 }, { "epoch": 7.18, "learning_rate": 7.505501100220045e-07, "loss": 0.0004, "step": 23133 }, { "epoch": 7.18, "learning_rate": 7.501500300060012e-07, "loss": 0.0017, "step": 23134 }, { "epoch": 7.18, "learning_rate": 7.497499499899981e-07, "loss": 0.0003, "step": 23135 }, { "epoch": 7.18, "learning_rate": 7.493498699739948e-07, "loss": 0.0024, "step": 23136 }, { "epoch": 7.18, "learning_rate": 7.489497899579917e-07, "loss": 0.0006, "step": 23137 }, { "epoch": 7.18, "learning_rate": 7.485497099419884e-07, "loss": 0.0009, "step": 23138 }, { "epoch": 7.18, "learning_rate": 7.481496299259852e-07, "loss": 0.0013, "step": 23139 }, { "epoch": 7.18, "learning_rate": 7.477495499099821e-07, "loss": 0.0132, "step": 23140 }, { "epoch": 7.18, "learning_rate": 7.473494698939788e-07, "loss": 0.0007, "step": 23141 }, { "epoch": 7.18, "learning_rate": 7.469493898779756e-07, "loss": 0.0012, "step": 23142 }, { "epoch": 7.18, "learning_rate": 7.465493098619725e-07, "loss": 0.0006, "step": 23143 }, { "epoch": 7.18, "learning_rate": 7.461492298459693e-07, "loss": 0.0012, "step": 23144 }, { "epoch": 7.18, "learning_rate": 7.45749149829966e-07, "loss": 0.0026, "step": 23145 }, { "epoch": 7.18, "learning_rate": 7.453490698139629e-07, "loss": 0.0007, "step": 23146 }, { "epoch": 7.18, "learning_rate": 7.449489897979597e-07, "loss": 0.0004, "step": 23147 }, { "epoch": 7.18, "learning_rate": 7.445489097819566e-07, "loss": 0.0006, "step": 23148 }, { "epoch": 7.18, "learning_rate": 7.441488297659532e-07, "loss": 0.0037, "step": 23149 }, { "epoch": 7.19, "learning_rate": 7.4374874974995e-07, "loss": 0.0009, "step": 23150 }, { "epoch": 7.19, "learning_rate": 7.433486697339469e-07, "loss": 0.0011, "step": 23151 }, { "epoch": 7.19, "learning_rate": 7.429485897179436e-07, "loss": 0.0004, "step": 23152 }, { "epoch": 7.19, "learning_rate": 7.425485097019404e-07, "loss": 0.0165, "step": 23153 }, { "epoch": 7.19, "learning_rate": 7.421484296859373e-07, "loss": 0.0029, "step": 23154 }, { "epoch": 7.19, "learning_rate": 7.41748349669934e-07, "loss": 0.0017, "step": 23155 }, { "epoch": 7.19, "learning_rate": 7.413482696539308e-07, "loss": 0.0007, "step": 23156 }, { "epoch": 7.19, "learning_rate": 7.409481896379277e-07, "loss": 0.0009, "step": 23157 }, { "epoch": 7.19, "learning_rate": 7.405481096219244e-07, "loss": 0.0008, "step": 23158 }, { "epoch": 7.19, "learning_rate": 7.401480296059213e-07, "loss": 0.0008, "step": 23159 }, { "epoch": 7.19, "learning_rate": 7.397479495899181e-07, "loss": 0.0146, "step": 23160 }, { "epoch": 7.19, "learning_rate": 7.393478695739148e-07, "loss": 0.0007, "step": 23161 }, { "epoch": 7.19, "learning_rate": 7.389477895579117e-07, "loss": 0.0008, "step": 23162 }, { "epoch": 7.19, "learning_rate": 7.385477095419084e-07, "loss": 0.001, "step": 23163 }, { "epoch": 7.19, "learning_rate": 7.381476295259051e-07, "loss": 0.0009, "step": 23164 }, { "epoch": 7.19, "learning_rate": 7.37747549509902e-07, "loss": 0.002, "step": 23165 }, { "epoch": 7.19, "learning_rate": 7.373474694938988e-07, "loss": 0.0008, "step": 23166 }, { "epoch": 7.19, "learning_rate": 7.369473894778957e-07, "loss": 0.0012, "step": 23167 }, { "epoch": 7.19, "learning_rate": 7.365473094618924e-07, "loss": 0.0007, "step": 23168 }, { "epoch": 7.19, "learning_rate": 7.361472294458892e-07, "loss": 0.0008, "step": 23169 }, { "epoch": 7.19, "learning_rate": 7.357471494298861e-07, "loss": 0.0055, "step": 23170 }, { "epoch": 7.19, "learning_rate": 7.353470694138829e-07, "loss": 0.0009, "step": 23171 }, { "epoch": 7.19, "learning_rate": 7.349469893978796e-07, "loss": 0.0006, "step": 23172 }, { "epoch": 7.19, "learning_rate": 7.345469093818765e-07, "loss": 0.0156, "step": 23173 }, { "epoch": 7.19, "learning_rate": 7.341468293658733e-07, "loss": 0.0009, "step": 23174 }, { "epoch": 7.19, "learning_rate": 7.3374674934987e-07, "loss": 0.0005, "step": 23175 }, { "epoch": 7.19, "learning_rate": 7.333466693338668e-07, "loss": 0.0011, "step": 23176 }, { "epoch": 7.19, "learning_rate": 7.329465893178636e-07, "loss": 0.0015, "step": 23177 }, { "epoch": 7.19, "learning_rate": 7.325465093018605e-07, "loss": 0.0011, "step": 23178 }, { "epoch": 7.19, "learning_rate": 7.321464292858572e-07, "loss": 0.0008, "step": 23179 }, { "epoch": 7.19, "learning_rate": 7.31746349269854e-07, "loss": 0.0009, "step": 23180 }, { "epoch": 7.19, "learning_rate": 7.313462692538509e-07, "loss": 0.0007, "step": 23181 }, { "epoch": 7.2, "learning_rate": 7.309461892378476e-07, "loss": 0.0009, "step": 23182 }, { "epoch": 7.2, "learning_rate": 7.305461092218444e-07, "loss": 0.001, "step": 23183 }, { "epoch": 7.2, "learning_rate": 7.301460292058413e-07, "loss": 0.0037, "step": 23184 }, { "epoch": 7.2, "learning_rate": 7.29745949189838e-07, "loss": 0.0027, "step": 23185 }, { "epoch": 7.2, "learning_rate": 7.293458691738348e-07, "loss": 0.0011, "step": 23186 }, { "epoch": 7.2, "learning_rate": 7.289457891578317e-07, "loss": 0.0009, "step": 23187 }, { "epoch": 7.2, "learning_rate": 7.285457091418284e-07, "loss": 0.0006, "step": 23188 }, { "epoch": 7.2, "learning_rate": 7.281456291258253e-07, "loss": 0.0008, "step": 23189 }, { "epoch": 7.2, "learning_rate": 7.27745549109822e-07, "loss": 0.0005, "step": 23190 }, { "epoch": 7.2, "learning_rate": 7.273454690938187e-07, "loss": 0.0006, "step": 23191 }, { "epoch": 7.2, "learning_rate": 7.269453890778156e-07, "loss": 0.0008, "step": 23192 }, { "epoch": 7.2, "learning_rate": 7.265453090618124e-07, "loss": 0.0015, "step": 23193 }, { "epoch": 7.2, "learning_rate": 7.261452290458092e-07, "loss": 0.0007, "step": 23194 }, { "epoch": 7.2, "learning_rate": 7.25745149029806e-07, "loss": 0.0006, "step": 23195 }, { "epoch": 7.2, "learning_rate": 7.253450690138028e-07, "loss": 0.042, "step": 23196 }, { "epoch": 7.2, "learning_rate": 7.249449889977996e-07, "loss": 0.0015, "step": 23197 }, { "epoch": 7.2, "learning_rate": 7.245449089817965e-07, "loss": 0.0007, "step": 23198 }, { "epoch": 7.2, "learning_rate": 7.241448289657932e-07, "loss": 0.0095, "step": 23199 }, { "epoch": 7.2, "learning_rate": 7.237447489497901e-07, "loss": 0.0007, "step": 23200 }, { "epoch": 7.2, "learning_rate": 7.233446689337869e-07, "loss": 0.0007, "step": 23201 }, { "epoch": 7.2, "learning_rate": 7.229445889177836e-07, "loss": 0.0114, "step": 23202 }, { "epoch": 7.2, "learning_rate": 7.225445089017804e-07, "loss": 0.0113, "step": 23203 }, { "epoch": 7.2, "learning_rate": 7.221444288857772e-07, "loss": 0.001, "step": 23204 }, { "epoch": 7.2, "learning_rate": 7.217443488697739e-07, "loss": 0.0006, "step": 23205 }, { "epoch": 7.2, "learning_rate": 7.213442688537708e-07, "loss": 0.001, "step": 23206 }, { "epoch": 7.2, "learning_rate": 7.209441888377676e-07, "loss": 0.0008, "step": 23207 }, { "epoch": 7.2, "learning_rate": 7.205441088217643e-07, "loss": 0.0006, "step": 23208 }, { "epoch": 7.2, "learning_rate": 7.201440288057612e-07, "loss": 0.0006, "step": 23209 }, { "epoch": 7.2, "learning_rate": 7.19743948789758e-07, "loss": 0.0003, "step": 23210 }, { "epoch": 7.2, "learning_rate": 7.193438687737549e-07, "loss": 0.0006, "step": 23211 }, { "epoch": 7.2, "learning_rate": 7.189437887577516e-07, "loss": 0.0017, "step": 23212 }, { "epoch": 7.2, "learning_rate": 7.185437087417484e-07, "loss": 0.0014, "step": 23213 }, { "epoch": 7.21, "learning_rate": 7.181436287257453e-07, "loss": 0.001, "step": 23214 }, { "epoch": 7.21, "learning_rate": 7.17743548709742e-07, "loss": 0.0259, "step": 23215 }, { "epoch": 7.21, "learning_rate": 7.173434686937387e-07, "loss": 0.0008, "step": 23216 }, { "epoch": 7.21, "learning_rate": 7.169433886777356e-07, "loss": 0.0007, "step": 23217 }, { "epoch": 7.21, "learning_rate": 7.165433086617323e-07, "loss": 0.0005, "step": 23218 }, { "epoch": 7.21, "learning_rate": 7.161432286457292e-07, "loss": 0.0006, "step": 23219 }, { "epoch": 7.21, "learning_rate": 7.15743148629726e-07, "loss": 0.0008, "step": 23220 }, { "epoch": 7.21, "learning_rate": 7.153430686137228e-07, "loss": 0.0106, "step": 23221 }, { "epoch": 7.21, "learning_rate": 7.149429885977196e-07, "loss": 0.0014, "step": 23222 }, { "epoch": 7.21, "learning_rate": 7.145429085817164e-07, "loss": 0.0026, "step": 23223 }, { "epoch": 7.21, "learning_rate": 7.141428285657132e-07, "loss": 0.0064, "step": 23224 }, { "epoch": 7.21, "learning_rate": 7.137427485497101e-07, "loss": 0.0007, "step": 23225 }, { "epoch": 7.21, "learning_rate": 7.133426685337068e-07, "loss": 0.0005, "step": 23226 }, { "epoch": 7.21, "learning_rate": 7.129425885177036e-07, "loss": 0.0018, "step": 23227 }, { "epoch": 7.21, "learning_rate": 7.125425085017005e-07, "loss": 0.0009, "step": 23228 }, { "epoch": 7.21, "learning_rate": 7.121424284856972e-07, "loss": 0.0049, "step": 23229 }, { "epoch": 7.21, "learning_rate": 7.11742348469694e-07, "loss": 0.0021, "step": 23230 }, { "epoch": 7.21, "learning_rate": 7.113422684536908e-07, "loss": 0.0005, "step": 23231 }, { "epoch": 7.21, "learning_rate": 7.109421884376875e-07, "loss": 0.0008, "step": 23232 }, { "epoch": 7.21, "learning_rate": 7.105421084216844e-07, "loss": 0.0229, "step": 23233 }, { "epoch": 7.21, "learning_rate": 7.101420284056812e-07, "loss": 0.0008, "step": 23234 }, { "epoch": 7.21, "learning_rate": 7.097419483896779e-07, "loss": 0.0007, "step": 23235 }, { "epoch": 7.21, "learning_rate": 7.093418683736748e-07, "loss": 0.0091, "step": 23236 }, { "epoch": 7.21, "learning_rate": 7.089417883576716e-07, "loss": 0.0009, "step": 23237 }, { "epoch": 7.21, "learning_rate": 7.085417083416683e-07, "loss": 0.0003, "step": 23238 }, { "epoch": 7.21, "learning_rate": 7.081416283256652e-07, "loss": 0.0007, "step": 23239 }, { "epoch": 7.21, "learning_rate": 7.07741548309662e-07, "loss": 0.0011, "step": 23240 }, { "epoch": 7.21, "learning_rate": 7.073414682936589e-07, "loss": 0.0008, "step": 23241 }, { "epoch": 7.21, "learning_rate": 7.069413882776556e-07, "loss": 0.0007, "step": 23242 }, { "epoch": 7.21, "learning_rate": 7.065413082616523e-07, "loss": 0.0007, "step": 23243 }, { "epoch": 7.21, "learning_rate": 7.061412282456492e-07, "loss": 0.0013, "step": 23244 }, { "epoch": 7.21, "learning_rate": 7.057411482296459e-07, "loss": 0.0007, "step": 23245 }, { "epoch": 7.22, "learning_rate": 7.053410682136427e-07, "loss": 0.0005, "step": 23246 }, { "epoch": 7.22, "learning_rate": 7.049409881976396e-07, "loss": 0.0205, "step": 23247 }, { "epoch": 7.22, "learning_rate": 7.045409081816364e-07, "loss": 0.0007, "step": 23248 }, { "epoch": 7.22, "learning_rate": 7.041408281656331e-07, "loss": 0.0026, "step": 23249 }, { "epoch": 7.22, "learning_rate": 7.0374074814963e-07, "loss": 0.0075, "step": 23250 }, { "epoch": 7.22, "learning_rate": 7.033406681336268e-07, "loss": 0.0011, "step": 23251 }, { "epoch": 7.22, "learning_rate": 7.029405881176237e-07, "loss": 0.0007, "step": 23252 }, { "epoch": 7.22, "learning_rate": 7.025405081016204e-07, "loss": 0.0008, "step": 23253 }, { "epoch": 7.22, "learning_rate": 7.021404280856172e-07, "loss": 0.0008, "step": 23254 }, { "epoch": 7.22, "learning_rate": 7.017403480696141e-07, "loss": 0.0013, "step": 23255 }, { "epoch": 7.22, "learning_rate": 7.013402680536108e-07, "loss": 0.0019, "step": 23256 }, { "epoch": 7.22, "learning_rate": 7.009401880376075e-07, "loss": 0.0187, "step": 23257 }, { "epoch": 7.22, "learning_rate": 7.005401080216044e-07, "loss": 0.0013, "step": 23258 }, { "epoch": 7.22, "learning_rate": 7.001400280056011e-07, "loss": 0.0005, "step": 23259 }, { "epoch": 7.22, "learning_rate": 6.99739947989598e-07, "loss": 0.0007, "step": 23260 }, { "epoch": 7.22, "learning_rate": 6.993398679735948e-07, "loss": 0.0006, "step": 23261 }, { "epoch": 7.22, "learning_rate": 6.989397879575915e-07, "loss": 0.0009, "step": 23262 }, { "epoch": 7.22, "learning_rate": 6.985397079415884e-07, "loss": 0.0009, "step": 23263 }, { "epoch": 7.22, "learning_rate": 6.981396279255852e-07, "loss": 0.0009, "step": 23264 }, { "epoch": 7.22, "learning_rate": 6.977395479095819e-07, "loss": 0.0006, "step": 23265 }, { "epoch": 7.22, "learning_rate": 6.973394678935788e-07, "loss": 0.0009, "step": 23266 }, { "epoch": 7.22, "learning_rate": 6.969393878775756e-07, "loss": 0.0012, "step": 23267 }, { "epoch": 7.22, "learning_rate": 6.965393078615724e-07, "loss": 0.0006, "step": 23268 }, { "epoch": 7.22, "learning_rate": 6.961392278455692e-07, "loss": 0.0019, "step": 23269 }, { "epoch": 7.22, "learning_rate": 6.95739147829566e-07, "loss": 0.0006, "step": 23270 }, { "epoch": 7.22, "learning_rate": 6.953390678135628e-07, "loss": 0.0008, "step": 23271 }, { "epoch": 7.22, "learning_rate": 6.949389877975595e-07, "loss": 0.0006, "step": 23272 }, { "epoch": 7.22, "learning_rate": 6.945389077815563e-07, "loss": 0.0157, "step": 23273 }, { "epoch": 7.22, "learning_rate": 6.941388277655532e-07, "loss": 0.0007, "step": 23274 }, { "epoch": 7.22, "learning_rate": 6.9373874774955e-07, "loss": 0.0006, "step": 23275 }, { "epoch": 7.22, "learning_rate": 6.933386677335467e-07, "loss": 0.0011, "step": 23276 }, { "epoch": 7.22, "learning_rate": 6.929385877175436e-07, "loss": 0.0006, "step": 23277 }, { "epoch": 7.22, "learning_rate": 6.925385077015404e-07, "loss": 0.001, "step": 23278 }, { "epoch": 7.23, "learning_rate": 6.921384276855371e-07, "loss": 0.0009, "step": 23279 }, { "epoch": 7.23, "learning_rate": 6.91738347669534e-07, "loss": 0.0006, "step": 23280 }, { "epoch": 7.23, "learning_rate": 6.913382676535308e-07, "loss": 0.0007, "step": 23281 }, { "epoch": 7.23, "learning_rate": 6.909381876375277e-07, "loss": 0.0004, "step": 23282 }, { "epoch": 7.23, "learning_rate": 6.905381076215244e-07, "loss": 0.0017, "step": 23283 }, { "epoch": 7.23, "learning_rate": 6.901380276055211e-07, "loss": 0.0012, "step": 23284 }, { "epoch": 7.23, "learning_rate": 6.89737947589518e-07, "loss": 0.0044, "step": 23285 }, { "epoch": 7.23, "learning_rate": 6.893378675735147e-07, "loss": 0.0007, "step": 23286 }, { "epoch": 7.23, "learning_rate": 6.889377875575115e-07, "loss": 0.0006, "step": 23287 }, { "epoch": 7.23, "learning_rate": 6.885377075415084e-07, "loss": 0.0007, "step": 23288 }, { "epoch": 7.23, "learning_rate": 6.881376275255051e-07, "loss": 0.0006, "step": 23289 }, { "epoch": 7.23, "learning_rate": 6.877375475095019e-07, "loss": 0.0009, "step": 23290 }, { "epoch": 7.23, "learning_rate": 6.873374674934988e-07, "loss": 0.0006, "step": 23291 }, { "epoch": 7.23, "learning_rate": 6.869373874774955e-07, "loss": 0.0008, "step": 23292 }, { "epoch": 7.23, "learning_rate": 6.865373074614924e-07, "loss": 0.0004, "step": 23293 }, { "epoch": 7.23, "learning_rate": 6.861372274454892e-07, "loss": 0.0003, "step": 23294 }, { "epoch": 7.23, "learning_rate": 6.85737147429486e-07, "loss": 0.001, "step": 23295 }, { "epoch": 7.23, "learning_rate": 6.853370674134828e-07, "loss": 0.0007, "step": 23296 }, { "epoch": 7.23, "learning_rate": 6.849369873974795e-07, "loss": 0.0005, "step": 23297 }, { "epoch": 7.23, "learning_rate": 6.845369073814763e-07, "loss": 0.0019, "step": 23298 }, { "epoch": 7.23, "learning_rate": 6.841368273654731e-07, "loss": 0.0005, "step": 23299 }, { "epoch": 7.23, "learning_rate": 6.837367473494699e-07, "loss": 0.0008, "step": 23300 }, { "epoch": 7.23, "learning_rate": 6.833366673334667e-07, "loss": 0.0008, "step": 23301 }, { "epoch": 7.23, "learning_rate": 6.829365873174636e-07, "loss": 0.0037, "step": 23302 }, { "epoch": 7.23, "learning_rate": 6.825365073014603e-07, "loss": 0.0008, "step": 23303 }, { "epoch": 7.23, "learning_rate": 6.821364272854572e-07, "loss": 0.001, "step": 23304 }, { "epoch": 7.23, "learning_rate": 6.81736347269454e-07, "loss": 0.0169, "step": 23305 }, { "epoch": 7.23, "learning_rate": 6.813362672534507e-07, "loss": 0.0011, "step": 23306 }, { "epoch": 7.23, "learning_rate": 6.809361872374476e-07, "loss": 0.0006, "step": 23307 }, { "epoch": 7.23, "learning_rate": 6.805361072214444e-07, "loss": 0.0087, "step": 23308 }, { "epoch": 7.23, "learning_rate": 6.801360272054411e-07, "loss": 0.0005, "step": 23309 }, { "epoch": 7.23, "learning_rate": 6.79735947189438e-07, "loss": 0.0046, "step": 23310 }, { "epoch": 7.24, "learning_rate": 6.793358671734347e-07, "loss": 0.0005, "step": 23311 }, { "epoch": 7.24, "learning_rate": 6.789357871574316e-07, "loss": 0.0011, "step": 23312 }, { "epoch": 7.24, "learning_rate": 6.785357071414283e-07, "loss": 0.013, "step": 23313 }, { "epoch": 7.24, "learning_rate": 6.781356271254251e-07, "loss": 0.0008, "step": 23314 }, { "epoch": 7.24, "learning_rate": 6.77735547109422e-07, "loss": 0.0007, "step": 23315 }, { "epoch": 7.24, "learning_rate": 6.773354670934187e-07, "loss": 0.0007, "step": 23316 }, { "epoch": 7.24, "learning_rate": 6.769353870774155e-07, "loss": 0.0004, "step": 23317 }, { "epoch": 7.24, "learning_rate": 6.765353070614124e-07, "loss": 0.0007, "step": 23318 }, { "epoch": 7.24, "learning_rate": 6.761352270454091e-07, "loss": 0.0297, "step": 23319 }, { "epoch": 7.24, "learning_rate": 6.757351470294059e-07, "loss": 0.0009, "step": 23320 }, { "epoch": 7.24, "learning_rate": 6.753350670134028e-07, "loss": 0.0013, "step": 23321 }, { "epoch": 7.24, "learning_rate": 6.749349869973996e-07, "loss": 0.0006, "step": 23322 }, { "epoch": 7.24, "learning_rate": 6.745349069813964e-07, "loss": 0.0008, "step": 23323 }, { "epoch": 7.24, "learning_rate": 6.741348269653931e-07, "loss": 0.0006, "step": 23324 }, { "epoch": 7.24, "learning_rate": 6.737347469493899e-07, "loss": 0.0011, "step": 23325 }, { "epoch": 7.24, "learning_rate": 6.733346669333867e-07, "loss": 0.0005, "step": 23326 }, { "epoch": 7.24, "learning_rate": 6.729345869173835e-07, "loss": 0.0069, "step": 23327 }, { "epoch": 7.24, "learning_rate": 6.725345069013803e-07, "loss": 0.0006, "step": 23328 }, { "epoch": 7.24, "learning_rate": 6.721344268853772e-07, "loss": 0.0009, "step": 23329 }, { "epoch": 7.24, "learning_rate": 6.717343468693739e-07, "loss": 0.001, "step": 23330 }, { "epoch": 7.24, "learning_rate": 6.713342668533707e-07, "loss": 0.0006, "step": 23331 }, { "epoch": 7.24, "learning_rate": 6.709341868373676e-07, "loss": 0.0009, "step": 23332 }, { "epoch": 7.24, "learning_rate": 6.705341068213643e-07, "loss": 0.0008, "step": 23333 }, { "epoch": 7.24, "learning_rate": 6.701340268053612e-07, "loss": 0.0005, "step": 23334 }, { "epoch": 7.24, "learning_rate": 6.69733946789358e-07, "loss": 0.0007, "step": 23335 }, { "epoch": 7.24, "learning_rate": 6.693338667733547e-07, "loss": 0.0007, "step": 23336 }, { "epoch": 7.24, "learning_rate": 6.689337867573516e-07, "loss": 0.0011, "step": 23337 }, { "epoch": 7.24, "learning_rate": 6.685337067413483e-07, "loss": 0.0016, "step": 23338 }, { "epoch": 7.24, "learning_rate": 6.68133626725345e-07, "loss": 0.0021, "step": 23339 }, { "epoch": 7.24, "learning_rate": 6.677335467093419e-07, "loss": 0.0005, "step": 23340 }, { "epoch": 7.24, "learning_rate": 6.673334666933387e-07, "loss": 0.0188, "step": 23341 }, { "epoch": 7.24, "learning_rate": 6.669333866773354e-07, "loss": 0.0184, "step": 23342 }, { "epoch": 7.25, "learning_rate": 6.665333066613323e-07, "loss": 0.0004, "step": 23343 }, { "epoch": 7.25, "learning_rate": 6.661332266453291e-07, "loss": 0.0018, "step": 23344 }, { "epoch": 7.25, "learning_rate": 6.65733146629326e-07, "loss": 0.0011, "step": 23345 }, { "epoch": 7.25, "learning_rate": 6.653330666133227e-07, "loss": 0.0007, "step": 23346 }, { "epoch": 7.25, "learning_rate": 6.649329865973195e-07, "loss": 0.001, "step": 23347 }, { "epoch": 7.25, "learning_rate": 6.645329065813164e-07, "loss": 0.0009, "step": 23348 }, { "epoch": 7.25, "learning_rate": 6.641328265653131e-07, "loss": 0.0007, "step": 23349 }, { "epoch": 7.25, "learning_rate": 6.637327465493098e-07, "loss": 0.0024, "step": 23350 }, { "epoch": 7.25, "learning_rate": 6.633326665333067e-07, "loss": 0.0006, "step": 23351 }, { "epoch": 7.25, "learning_rate": 6.629325865173035e-07, "loss": 0.0007, "step": 23352 }, { "epoch": 7.25, "learning_rate": 6.625325065013003e-07, "loss": 0.0184, "step": 23353 }, { "epoch": 7.25, "learning_rate": 6.621324264852971e-07, "loss": 0.0015, "step": 23354 }, { "epoch": 7.25, "learning_rate": 6.617323464692939e-07, "loss": 0.0011, "step": 23355 }, { "epoch": 7.25, "learning_rate": 6.613322664532908e-07, "loss": 0.0008, "step": 23356 }, { "epoch": 7.25, "learning_rate": 6.609321864372875e-07, "loss": 0.0019, "step": 23357 }, { "epoch": 7.25, "learning_rate": 6.605321064212843e-07, "loss": 0.0007, "step": 23358 }, { "epoch": 7.25, "learning_rate": 6.601320264052812e-07, "loss": 0.0011, "step": 23359 }, { "epoch": 7.25, "learning_rate": 6.597319463892779e-07, "loss": 0.001, "step": 23360 }, { "epoch": 7.25, "learning_rate": 6.593318663732747e-07, "loss": 0.0009, "step": 23361 }, { "epoch": 7.25, "learning_rate": 6.589317863572716e-07, "loss": 0.0005, "step": 23362 }, { "epoch": 7.25, "learning_rate": 6.585317063412683e-07, "loss": 0.0006, "step": 23363 }, { "epoch": 7.25, "learning_rate": 6.581316263252652e-07, "loss": 0.001, "step": 23364 }, { "epoch": 7.25, "learning_rate": 6.577315463092619e-07, "loss": 0.0007, "step": 23365 }, { "epoch": 7.25, "learning_rate": 6.573314662932586e-07, "loss": 0.0188, "step": 23366 }, { "epoch": 7.25, "learning_rate": 6.569313862772555e-07, "loss": 0.0006, "step": 23367 }, { "epoch": 7.25, "learning_rate": 6.565313062612523e-07, "loss": 0.0015, "step": 23368 }, { "epoch": 7.25, "learning_rate": 6.56131226245249e-07, "loss": 0.001, "step": 23369 }, { "epoch": 7.25, "learning_rate": 6.557311462292459e-07, "loss": 0.0217, "step": 23370 }, { "epoch": 7.25, "learning_rate": 6.553310662132427e-07, "loss": 0.0011, "step": 23371 }, { "epoch": 7.25, "learning_rate": 6.549309861972395e-07, "loss": 0.001, "step": 23372 }, { "epoch": 7.25, "learning_rate": 6.545309061812363e-07, "loss": 0.003, "step": 23373 }, { "epoch": 7.25, "learning_rate": 6.541308261652331e-07, "loss": 0.0012, "step": 23374 }, { "epoch": 7.26, "learning_rate": 6.5373074614923e-07, "loss": 0.0008, "step": 23375 }, { "epoch": 7.26, "learning_rate": 6.533306661332267e-07, "loss": 0.0007, "step": 23376 }, { "epoch": 7.26, "learning_rate": 6.529305861172234e-07, "loss": 0.0008, "step": 23377 }, { "epoch": 7.26, "learning_rate": 6.525305061012203e-07, "loss": 0.0006, "step": 23378 }, { "epoch": 7.26, "learning_rate": 6.521304260852171e-07, "loss": 0.0023, "step": 23379 }, { "epoch": 7.26, "learning_rate": 6.517303460692138e-07, "loss": 0.0033, "step": 23380 }, { "epoch": 7.26, "learning_rate": 6.513302660532107e-07, "loss": 0.0012, "step": 23381 }, { "epoch": 7.26, "learning_rate": 6.509301860372075e-07, "loss": 0.0008, "step": 23382 }, { "epoch": 7.26, "learning_rate": 6.505301060212042e-07, "loss": 0.0005, "step": 23383 }, { "epoch": 7.26, "learning_rate": 6.501300260052011e-07, "loss": 0.001, "step": 23384 }, { "epoch": 7.26, "learning_rate": 6.497299459891979e-07, "loss": 0.0008, "step": 23385 }, { "epoch": 7.26, "learning_rate": 6.493298659731948e-07, "loss": 0.0005, "step": 23386 }, { "epoch": 7.26, "learning_rate": 6.489297859571915e-07, "loss": 0.0009, "step": 23387 }, { "epoch": 7.26, "learning_rate": 6.485297059411883e-07, "loss": 0.0006, "step": 23388 }, { "epoch": 7.26, "learning_rate": 6.481296259251852e-07, "loss": 0.0013, "step": 23389 }, { "epoch": 7.26, "learning_rate": 6.477295459091819e-07, "loss": 0.0014, "step": 23390 }, { "epoch": 7.26, "learning_rate": 6.473294658931786e-07, "loss": 0.0012, "step": 23391 }, { "epoch": 7.26, "learning_rate": 6.469293858771755e-07, "loss": 0.0013, "step": 23392 }, { "epoch": 7.26, "learning_rate": 6.465293058611722e-07, "loss": 0.0257, "step": 23393 }, { "epoch": 7.26, "learning_rate": 6.461292258451691e-07, "loss": 0.0008, "step": 23394 }, { "epoch": 7.26, "learning_rate": 6.457291458291659e-07, "loss": 0.0008, "step": 23395 }, { "epoch": 7.26, "learning_rate": 6.453290658131626e-07, "loss": 0.0006, "step": 23396 }, { "epoch": 7.26, "learning_rate": 6.449289857971595e-07, "loss": 0.001, "step": 23397 }, { "epoch": 7.26, "learning_rate": 6.445289057811563e-07, "loss": 0.0006, "step": 23398 }, { "epoch": 7.26, "learning_rate": 6.441288257651531e-07, "loss": 0.0005, "step": 23399 }, { "epoch": 7.26, "learning_rate": 6.437287457491499e-07, "loss": 0.0182, "step": 23400 }, { "epoch": 7.26, "learning_rate": 6.433286657331467e-07, "loss": 0.0006, "step": 23401 }, { "epoch": 7.26, "learning_rate": 6.429285857171435e-07, "loss": 0.0009, "step": 23402 }, { "epoch": 7.26, "learning_rate": 6.425285057011403e-07, "loss": 0.0009, "step": 23403 }, { "epoch": 7.26, "learning_rate": 6.42128425685137e-07, "loss": 0.0011, "step": 23404 }, { "epoch": 7.26, "learning_rate": 6.41728345669134e-07, "loss": 0.0047, "step": 23405 }, { "epoch": 7.26, "learning_rate": 6.413282656531307e-07, "loss": 0.0006, "step": 23406 }, { "epoch": 7.27, "learning_rate": 6.409281856371274e-07, "loss": 0.0008, "step": 23407 }, { "epoch": 7.27, "learning_rate": 6.405281056211243e-07, "loss": 0.0011, "step": 23408 }, { "epoch": 7.27, "learning_rate": 6.401280256051211e-07, "loss": 0.0007, "step": 23409 }, { "epoch": 7.27, "learning_rate": 6.397279455891178e-07, "loss": 0.0012, "step": 23410 }, { "epoch": 7.27, "learning_rate": 6.393278655731147e-07, "loss": 0.0007, "step": 23411 }, { "epoch": 7.27, "learning_rate": 6.389277855571115e-07, "loss": 0.001, "step": 23412 }, { "epoch": 7.27, "learning_rate": 6.385277055411082e-07, "loss": 0.0019, "step": 23413 }, { "epoch": 7.27, "learning_rate": 6.381276255251051e-07, "loss": 0.0016, "step": 23414 }, { "epoch": 7.27, "learning_rate": 6.377275455091019e-07, "loss": 0.0008, "step": 23415 }, { "epoch": 7.27, "learning_rate": 6.373274654930988e-07, "loss": 0.001, "step": 23416 }, { "epoch": 7.27, "learning_rate": 6.369273854770955e-07, "loss": 0.0009, "step": 23417 }, { "epoch": 7.27, "learning_rate": 6.365273054610922e-07, "loss": 0.0011, "step": 23418 }, { "epoch": 7.27, "learning_rate": 6.361272254450891e-07, "loss": 0.0013, "step": 23419 }, { "epoch": 7.27, "learning_rate": 6.357271454290858e-07, "loss": 0.0014, "step": 23420 }, { "epoch": 7.27, "learning_rate": 6.353270654130826e-07, "loss": 0.0006, "step": 23421 }, { "epoch": 7.27, "learning_rate": 6.349269853970795e-07, "loss": 0.0009, "step": 23422 }, { "epoch": 7.27, "learning_rate": 6.345269053810762e-07, "loss": 0.0007, "step": 23423 }, { "epoch": 7.27, "learning_rate": 6.34126825365073e-07, "loss": 0.001, "step": 23424 }, { "epoch": 7.27, "learning_rate": 6.337267453490699e-07, "loss": 0.0006, "step": 23425 }, { "epoch": 7.27, "learning_rate": 6.333266653330667e-07, "loss": 0.0028, "step": 23426 }, { "epoch": 7.27, "learning_rate": 6.329265853170635e-07, "loss": 0.0005, "step": 23427 }, { "epoch": 7.27, "learning_rate": 6.325265053010603e-07, "loss": 0.0008, "step": 23428 }, { "epoch": 7.27, "learning_rate": 6.321264252850571e-07, "loss": 0.0007, "step": 23429 }, { "epoch": 7.27, "learning_rate": 6.317263452690539e-07, "loss": 0.0019, "step": 23430 }, { "epoch": 7.27, "learning_rate": 6.313262652530506e-07, "loss": 0.0005, "step": 23431 }, { "epoch": 7.27, "learning_rate": 6.309261852370474e-07, "loss": 0.0294, "step": 23432 }, { "epoch": 7.27, "learning_rate": 6.305261052210443e-07, "loss": 0.0008, "step": 23433 }, { "epoch": 7.27, "learning_rate": 6.30126025205041e-07, "loss": 0.0066, "step": 23434 }, { "epoch": 7.27, "learning_rate": 6.297259451890378e-07, "loss": 0.0007, "step": 23435 }, { "epoch": 7.27, "learning_rate": 6.293258651730347e-07, "loss": 0.0005, "step": 23436 }, { "epoch": 7.27, "learning_rate": 6.289257851570314e-07, "loss": 0.0007, "step": 23437 }, { "epoch": 7.27, "learning_rate": 6.285257051410283e-07, "loss": 0.0015, "step": 23438 }, { "epoch": 7.27, "learning_rate": 6.281256251250251e-07, "loss": 0.0005, "step": 23439 }, { "epoch": 7.28, "learning_rate": 6.277255451090218e-07, "loss": 0.0006, "step": 23440 }, { "epoch": 7.28, "learning_rate": 6.273254650930187e-07, "loss": 0.0008, "step": 23441 }, { "epoch": 7.28, "learning_rate": 6.269253850770155e-07, "loss": 0.0005, "step": 23442 }, { "epoch": 7.28, "learning_rate": 6.265253050610122e-07, "loss": 0.0011, "step": 23443 }, { "epoch": 7.28, "learning_rate": 6.261252250450091e-07, "loss": 0.0007, "step": 23444 }, { "epoch": 7.28, "learning_rate": 6.257251450290058e-07, "loss": 0.0009, "step": 23445 }, { "epoch": 7.28, "learning_rate": 6.253250650130027e-07, "loss": 0.0025, "step": 23446 }, { "epoch": 7.28, "learning_rate": 6.249249849969994e-07, "loss": 0.004, "step": 23447 }, { "epoch": 7.28, "learning_rate": 6.245249049809962e-07, "loss": 0.0019, "step": 23448 }, { "epoch": 7.28, "learning_rate": 6.24124824964993e-07, "loss": 0.0005, "step": 23449 }, { "epoch": 7.28, "learning_rate": 6.237247449489898e-07, "loss": 0.0009, "step": 23450 }, { "epoch": 7.28, "learning_rate": 6.233246649329867e-07, "loss": 0.0007, "step": 23451 }, { "epoch": 7.28, "learning_rate": 6.229245849169834e-07, "loss": 0.0198, "step": 23452 }, { "epoch": 7.28, "learning_rate": 6.225245049009803e-07, "loss": 0.0013, "step": 23453 }, { "epoch": 7.28, "learning_rate": 6.221244248849771e-07, "loss": 0.0008, "step": 23454 }, { "epoch": 7.28, "learning_rate": 6.217243448689739e-07, "loss": 0.019, "step": 23455 }, { "epoch": 7.28, "learning_rate": 6.213242648529707e-07, "loss": 0.001, "step": 23456 }, { "epoch": 7.28, "learning_rate": 6.209241848369675e-07, "loss": 0.0012, "step": 23457 }, { "epoch": 7.28, "learning_rate": 6.205241048209642e-07, "loss": 0.0008, "step": 23458 }, { "epoch": 7.28, "learning_rate": 6.20124024804961e-07, "loss": 0.0018, "step": 23459 }, { "epoch": 7.28, "learning_rate": 6.197239447889578e-07, "loss": 0.0011, "step": 23460 }, { "epoch": 7.28, "learning_rate": 6.193238647729546e-07, "loss": 0.0009, "step": 23461 }, { "epoch": 7.28, "learning_rate": 6.189237847569514e-07, "loss": 0.0005, "step": 23462 }, { "epoch": 7.28, "learning_rate": 6.185237047409482e-07, "loss": 0.0006, "step": 23463 }, { "epoch": 7.28, "learning_rate": 6.18123624724945e-07, "loss": 0.0008, "step": 23464 }, { "epoch": 7.28, "learning_rate": 6.177235447089418e-07, "loss": 0.0007, "step": 23465 }, { "epoch": 7.28, "learning_rate": 6.173234646929387e-07, "loss": 0.0006, "step": 23466 }, { "epoch": 7.28, "learning_rate": 6.169233846769354e-07, "loss": 0.0005, "step": 23467 }, { "epoch": 7.28, "learning_rate": 6.165233046609323e-07, "loss": 0.0006, "step": 23468 }, { "epoch": 7.28, "learning_rate": 6.161232246449291e-07, "loss": 0.0003, "step": 23469 }, { "epoch": 7.28, "learning_rate": 6.157231446289259e-07, "loss": 0.0006, "step": 23470 }, { "epoch": 7.28, "learning_rate": 6.153230646129227e-07, "loss": 0.001, "step": 23471 }, { "epoch": 7.29, "learning_rate": 6.149229845969194e-07, "loss": 0.0026, "step": 23472 }, { "epoch": 7.29, "learning_rate": 6.145229045809162e-07, "loss": 0.0163, "step": 23473 }, { "epoch": 7.29, "learning_rate": 6.14122824564913e-07, "loss": 0.0008, "step": 23474 }, { "epoch": 7.29, "learning_rate": 6.137227445489098e-07, "loss": 0.0005, "step": 23475 }, { "epoch": 7.29, "learning_rate": 6.133226645329066e-07, "loss": 0.0008, "step": 23476 }, { "epoch": 7.29, "learning_rate": 6.129225845169034e-07, "loss": 0.0014, "step": 23477 }, { "epoch": 7.29, "learning_rate": 6.125225045009002e-07, "loss": 0.0011, "step": 23478 }, { "epoch": 7.29, "learning_rate": 6.12122424484897e-07, "loss": 0.0007, "step": 23479 }, { "epoch": 7.29, "learning_rate": 6.117223444688939e-07, "loss": 0.0027, "step": 23480 }, { "epoch": 7.29, "learning_rate": 6.113222644528907e-07, "loss": 0.0014, "step": 23481 }, { "epoch": 7.29, "learning_rate": 6.109221844368874e-07, "loss": 0.0005, "step": 23482 }, { "epoch": 7.29, "learning_rate": 6.105221044208843e-07, "loss": 0.002, "step": 23483 }, { "epoch": 7.29, "learning_rate": 6.101220244048811e-07, "loss": 0.0007, "step": 23484 }, { "epoch": 7.29, "learning_rate": 6.097219443888778e-07, "loss": 0.0029, "step": 23485 }, { "epoch": 7.29, "learning_rate": 6.093218643728746e-07, "loss": 0.0009, "step": 23486 }, { "epoch": 7.29, "learning_rate": 6.089217843568714e-07, "loss": 0.0017, "step": 23487 }, { "epoch": 7.29, "learning_rate": 6.085217043408682e-07, "loss": 0.0007, "step": 23488 }, { "epoch": 7.29, "learning_rate": 6.08121624324865e-07, "loss": 0.0365, "step": 23489 }, { "epoch": 7.29, "learning_rate": 6.077215443088618e-07, "loss": 0.0025, "step": 23490 }, { "epoch": 7.29, "learning_rate": 6.073214642928586e-07, "loss": 0.0018, "step": 23491 }, { "epoch": 7.29, "learning_rate": 6.069213842768554e-07, "loss": 0.0007, "step": 23492 }, { "epoch": 7.29, "learning_rate": 6.065213042608522e-07, "loss": 0.0009, "step": 23493 }, { "epoch": 7.29, "learning_rate": 6.06121224244849e-07, "loss": 0.0006, "step": 23494 }, { "epoch": 7.29, "learning_rate": 6.057211442288459e-07, "loss": 0.0007, "step": 23495 }, { "epoch": 7.29, "learning_rate": 6.053210642128427e-07, "loss": 0.0008, "step": 23496 }, { "epoch": 7.29, "learning_rate": 6.049209841968394e-07, "loss": 0.001, "step": 23497 }, { "epoch": 7.29, "learning_rate": 6.045209041808363e-07, "loss": 0.0006, "step": 23498 }, { "epoch": 7.29, "learning_rate": 6.04120824164833e-07, "loss": 0.001, "step": 23499 }, { "epoch": 7.29, "learning_rate": 6.037207441488298e-07, "loss": 0.0009, "step": 23500 }, { "epoch": 7.29, "learning_rate": 6.033206641328266e-07, "loss": 0.014, "step": 23501 }, { "epoch": 7.29, "learning_rate": 6.029205841168234e-07, "loss": 0.0031, "step": 23502 }, { "epoch": 7.29, "learning_rate": 6.025205041008202e-07, "loss": 0.0008, "step": 23503 }, { "epoch": 7.3, "learning_rate": 6.02120424084817e-07, "loss": 0.0012, "step": 23504 }, { "epoch": 7.3, "learning_rate": 6.017203440688138e-07, "loss": 0.0009, "step": 23505 }, { "epoch": 7.3, "learning_rate": 6.013202640528106e-07, "loss": 0.0013, "step": 23506 }, { "epoch": 7.3, "learning_rate": 6.009201840368075e-07, "loss": 0.0005, "step": 23507 }, { "epoch": 7.3, "learning_rate": 6.005201040208042e-07, "loss": 0.0006, "step": 23508 }, { "epoch": 7.3, "learning_rate": 6.00120024004801e-07, "loss": 0.0262, "step": 23509 }, { "epoch": 7.3, "learning_rate": 5.997199439887979e-07, "loss": 0.0013, "step": 23510 }, { "epoch": 7.3, "learning_rate": 5.993198639727947e-07, "loss": 0.0009, "step": 23511 }, { "epoch": 7.3, "learning_rate": 5.989197839567913e-07, "loss": 0.0111, "step": 23512 }, { "epoch": 7.3, "learning_rate": 5.985197039407882e-07, "loss": 0.0024, "step": 23513 }, { "epoch": 7.3, "learning_rate": 5.98119623924785e-07, "loss": 0.0005, "step": 23514 }, { "epoch": 7.3, "learning_rate": 5.977195439087817e-07, "loss": 0.0005, "step": 23515 }, { "epoch": 7.3, "learning_rate": 5.973194638927786e-07, "loss": 0.0016, "step": 23516 }, { "epoch": 7.3, "learning_rate": 5.969193838767754e-07, "loss": 0.0008, "step": 23517 }, { "epoch": 7.3, "learning_rate": 5.965193038607722e-07, "loss": 0.0005, "step": 23518 }, { "epoch": 7.3, "learning_rate": 5.96119223844769e-07, "loss": 0.001, "step": 23519 }, { "epoch": 7.3, "learning_rate": 5.957191438287658e-07, "loss": 0.0005, "step": 23520 }, { "epoch": 7.3, "learning_rate": 5.953190638127626e-07, "loss": 0.001, "step": 23521 }, { "epoch": 7.3, "learning_rate": 5.949189837967595e-07, "loss": 0.0006, "step": 23522 }, { "epoch": 7.3, "learning_rate": 5.945189037807562e-07, "loss": 0.0007, "step": 23523 }, { "epoch": 7.3, "learning_rate": 5.94118823764753e-07, "loss": 0.0014, "step": 23524 }, { "epoch": 7.3, "learning_rate": 5.937187437487499e-07, "loss": 0.0008, "step": 23525 }, { "epoch": 7.3, "learning_rate": 5.933186637327466e-07, "loss": 0.0009, "step": 23526 }, { "epoch": 7.3, "learning_rate": 5.929185837167433e-07, "loss": 0.0011, "step": 23527 }, { "epoch": 7.3, "learning_rate": 5.925185037007402e-07, "loss": 0.0008, "step": 23528 }, { "epoch": 7.3, "learning_rate": 5.92118423684737e-07, "loss": 0.0011, "step": 23529 }, { "epoch": 7.3, "learning_rate": 5.917183436687338e-07, "loss": 0.0003, "step": 23530 }, { "epoch": 7.3, "learning_rate": 5.913182636527306e-07, "loss": 0.0005, "step": 23531 }, { "epoch": 7.3, "learning_rate": 5.909181836367274e-07, "loss": 0.0012, "step": 23532 }, { "epoch": 7.3, "learning_rate": 5.905181036207242e-07, "loss": 0.0007, "step": 23533 }, { "epoch": 7.3, "learning_rate": 5.90118023604721e-07, "loss": 0.0009, "step": 23534 }, { "epoch": 7.3, "learning_rate": 5.897179435887178e-07, "loss": 0.0008, "step": 23535 }, { "epoch": 7.31, "learning_rate": 5.893178635727146e-07, "loss": 0.0008, "step": 23536 }, { "epoch": 7.31, "learning_rate": 5.889177835567115e-07, "loss": 0.0011, "step": 23537 }, { "epoch": 7.31, "learning_rate": 5.885177035407082e-07, "loss": 0.001, "step": 23538 }, { "epoch": 7.31, "learning_rate": 5.881176235247049e-07, "loss": 0.0006, "step": 23539 }, { "epoch": 7.31, "learning_rate": 5.877175435087018e-07, "loss": 0.0006, "step": 23540 }, { "epoch": 7.31, "learning_rate": 5.873174634926986e-07, "loss": 0.0011, "step": 23541 }, { "epoch": 7.31, "learning_rate": 5.869173834766953e-07, "loss": 0.0022, "step": 23542 }, { "epoch": 7.31, "learning_rate": 5.865173034606922e-07, "loss": 0.0006, "step": 23543 }, { "epoch": 7.31, "learning_rate": 5.86117223444689e-07, "loss": 0.001, "step": 23544 }, { "epoch": 7.31, "learning_rate": 5.857171434286858e-07, "loss": 0.0004, "step": 23545 }, { "epoch": 7.31, "learning_rate": 5.853170634126826e-07, "loss": 0.0007, "step": 23546 }, { "epoch": 7.31, "learning_rate": 5.849169833966794e-07, "loss": 0.0008, "step": 23547 }, { "epoch": 7.31, "learning_rate": 5.845169033806762e-07, "loss": 0.007, "step": 23548 }, { "epoch": 7.31, "learning_rate": 5.84116823364673e-07, "loss": 0.0014, "step": 23549 }, { "epoch": 7.31, "learning_rate": 5.837167433486698e-07, "loss": 0.0005, "step": 23550 }, { "epoch": 7.31, "learning_rate": 5.833166633326666e-07, "loss": 0.0004, "step": 23551 }, { "epoch": 7.31, "learning_rate": 5.829165833166635e-07, "loss": 0.0011, "step": 23552 }, { "epoch": 7.31, "learning_rate": 5.825165033006601e-07, "loss": 0.024, "step": 23553 }, { "epoch": 7.31, "learning_rate": 5.821164232846569e-07, "loss": 0.0006, "step": 23554 }, { "epoch": 7.31, "learning_rate": 5.817163432686538e-07, "loss": 0.001, "step": 23555 }, { "epoch": 7.31, "learning_rate": 5.813162632526505e-07, "loss": 0.0004, "step": 23556 }, { "epoch": 7.31, "learning_rate": 5.809161832366474e-07, "loss": 0.0012, "step": 23557 }, { "epoch": 7.31, "learning_rate": 5.805161032206442e-07, "loss": 0.0006, "step": 23558 }, { "epoch": 7.31, "learning_rate": 5.80116023204641e-07, "loss": 0.0008, "step": 23559 }, { "epoch": 7.31, "learning_rate": 5.797159431886378e-07, "loss": 0.0009, "step": 23560 }, { "epoch": 7.31, "learning_rate": 5.793158631726346e-07, "loss": 0.0006, "step": 23561 }, { "epoch": 7.31, "learning_rate": 5.789157831566314e-07, "loss": 0.0005, "step": 23562 }, { "epoch": 7.31, "learning_rate": 5.785157031406282e-07, "loss": 0.0011, "step": 23563 }, { "epoch": 7.31, "learning_rate": 5.78115623124625e-07, "loss": 0.001, "step": 23564 }, { "epoch": 7.31, "learning_rate": 5.777155431086218e-07, "loss": 0.0006, "step": 23565 }, { "epoch": 7.31, "learning_rate": 5.773154630926185e-07, "loss": 0.0007, "step": 23566 }, { "epoch": 7.31, "learning_rate": 5.769153830766154e-07, "loss": 0.003, "step": 23567 }, { "epoch": 7.31, "learning_rate": 5.765153030606121e-07, "loss": 0.0012, "step": 23568 }, { "epoch": 7.32, "learning_rate": 5.761152230446089e-07, "loss": 0.0003, "step": 23569 }, { "epoch": 7.32, "learning_rate": 5.757151430286058e-07, "loss": 0.0005, "step": 23570 }, { "epoch": 7.32, "learning_rate": 5.753150630126025e-07, "loss": 0.0008, "step": 23571 }, { "epoch": 7.32, "learning_rate": 5.749149829965994e-07, "loss": 0.0047, "step": 23572 }, { "epoch": 7.32, "learning_rate": 5.745149029805962e-07, "loss": 0.0004, "step": 23573 }, { "epoch": 7.32, "learning_rate": 5.74114822964593e-07, "loss": 0.0011, "step": 23574 }, { "epoch": 7.32, "learning_rate": 5.737147429485898e-07, "loss": 0.0012, "step": 23575 }, { "epoch": 7.32, "learning_rate": 5.733146629325866e-07, "loss": 0.0008, "step": 23576 }, { "epoch": 7.32, "learning_rate": 5.729145829165834e-07, "loss": 0.001, "step": 23577 }, { "epoch": 7.32, "learning_rate": 5.725145029005802e-07, "loss": 0.0009, "step": 23578 }, { "epoch": 7.32, "learning_rate": 5.72114422884577e-07, "loss": 0.001, "step": 23579 }, { "epoch": 7.32, "learning_rate": 5.717143428685737e-07, "loss": 0.0006, "step": 23580 }, { "epoch": 7.32, "learning_rate": 5.713142628525705e-07, "loss": 0.0004, "step": 23581 }, { "epoch": 7.32, "learning_rate": 5.709141828365673e-07, "loss": 0.0007, "step": 23582 }, { "epoch": 7.32, "learning_rate": 5.705141028205641e-07, "loss": 0.0006, "step": 23583 }, { "epoch": 7.32, "learning_rate": 5.70114022804561e-07, "loss": 0.0007, "step": 23584 }, { "epoch": 7.32, "learning_rate": 5.697139427885578e-07, "loss": 0.0006, "step": 23585 }, { "epoch": 7.32, "learning_rate": 5.693138627725545e-07, "loss": 0.0034, "step": 23586 }, { "epoch": 7.32, "learning_rate": 5.689137827565514e-07, "loss": 0.0011, "step": 23587 }, { "epoch": 7.32, "learning_rate": 5.685137027405482e-07, "loss": 0.0007, "step": 23588 }, { "epoch": 7.32, "learning_rate": 5.68113622724545e-07, "loss": 0.0007, "step": 23589 }, { "epoch": 7.32, "learning_rate": 5.677135427085418e-07, "loss": 0.0004, "step": 23590 }, { "epoch": 7.32, "learning_rate": 5.673134626925386e-07, "loss": 0.0021, "step": 23591 }, { "epoch": 7.32, "learning_rate": 5.669133826765354e-07, "loss": 0.0051, "step": 23592 }, { "epoch": 7.32, "learning_rate": 5.665133026605321e-07, "loss": 0.001, "step": 23593 }, { "epoch": 7.32, "learning_rate": 5.661132226445289e-07, "loss": 0.0015, "step": 23594 }, { "epoch": 7.32, "learning_rate": 5.657131426285257e-07, "loss": 0.0016, "step": 23595 }, { "epoch": 7.32, "learning_rate": 5.653130626125225e-07, "loss": 0.0007, "step": 23596 }, { "epoch": 7.32, "learning_rate": 5.649129825965193e-07, "loss": 0.0009, "step": 23597 }, { "epoch": 7.32, "learning_rate": 5.645129025805161e-07, "loss": 0.0009, "step": 23598 }, { "epoch": 7.32, "learning_rate": 5.64112822564513e-07, "loss": 0.0008, "step": 23599 }, { "epoch": 7.32, "learning_rate": 5.637127425485098e-07, "loss": 0.0024, "step": 23600 }, { "epoch": 7.33, "learning_rate": 5.633126625325065e-07, "loss": 0.0003, "step": 23601 }, { "epoch": 7.33, "learning_rate": 5.629125825165034e-07, "loss": 0.0006, "step": 23602 }, { "epoch": 7.33, "learning_rate": 5.625125025005002e-07, "loss": 0.0007, "step": 23603 }, { "epoch": 7.33, "learning_rate": 5.62112422484497e-07, "loss": 0.001, "step": 23604 }, { "epoch": 7.33, "learning_rate": 5.617123424684938e-07, "loss": 0.0006, "step": 23605 }, { "epoch": 7.33, "learning_rate": 5.613122624524905e-07, "loss": 0.0014, "step": 23606 }, { "epoch": 7.33, "learning_rate": 5.609121824364873e-07, "loss": 0.0011, "step": 23607 }, { "epoch": 7.33, "learning_rate": 5.605121024204841e-07, "loss": 0.0007, "step": 23608 }, { "epoch": 7.33, "learning_rate": 5.601120224044809e-07, "loss": 0.0016, "step": 23609 }, { "epoch": 7.33, "learning_rate": 5.597119423884777e-07, "loss": 0.001, "step": 23610 }, { "epoch": 7.33, "learning_rate": 5.593118623724746e-07, "loss": 0.001, "step": 23611 }, { "epoch": 7.33, "learning_rate": 5.589117823564713e-07, "loss": 0.0005, "step": 23612 }, { "epoch": 7.33, "learning_rate": 5.585117023404681e-07, "loss": 0.001, "step": 23613 }, { "epoch": 7.33, "learning_rate": 5.58111622324465e-07, "loss": 0.0007, "step": 23614 }, { "epoch": 7.33, "learning_rate": 5.577115423084618e-07, "loss": 0.0006, "step": 23615 }, { "epoch": 7.33, "learning_rate": 5.573114622924585e-07, "loss": 0.0041, "step": 23616 }, { "epoch": 7.33, "learning_rate": 5.569113822764554e-07, "loss": 0.0017, "step": 23617 }, { "epoch": 7.33, "learning_rate": 5.565113022604522e-07, "loss": 0.0006, "step": 23618 }, { "epoch": 7.33, "learning_rate": 5.56111222244449e-07, "loss": 0.0029, "step": 23619 }, { "epoch": 7.33, "learning_rate": 5.557111422284457e-07, "loss": 0.0009, "step": 23620 }, { "epoch": 7.33, "learning_rate": 5.553110622124425e-07, "loss": 0.0008, "step": 23621 }, { "epoch": 7.33, "learning_rate": 5.549109821964393e-07, "loss": 0.0006, "step": 23622 }, { "epoch": 7.33, "learning_rate": 5.545109021804361e-07, "loss": 0.0009, "step": 23623 }, { "epoch": 7.33, "learning_rate": 5.541108221644329e-07, "loss": 0.0009, "step": 23624 }, { "epoch": 7.33, "learning_rate": 5.537107421484297e-07, "loss": 0.0013, "step": 23625 }, { "epoch": 7.33, "learning_rate": 5.533106621324266e-07, "loss": 0.0012, "step": 23626 }, { "epoch": 7.33, "learning_rate": 5.529105821164233e-07, "loss": 0.0014, "step": 23627 }, { "epoch": 7.33, "learning_rate": 5.525105021004201e-07, "loss": 0.0012, "step": 23628 }, { "epoch": 7.33, "learning_rate": 5.52110422084417e-07, "loss": 0.0012, "step": 23629 }, { "epoch": 7.33, "learning_rate": 5.517103420684138e-07, "loss": 0.001, "step": 23630 }, { "epoch": 7.33, "learning_rate": 5.513102620524105e-07, "loss": 0.0093, "step": 23631 }, { "epoch": 7.33, "learning_rate": 5.509101820364074e-07, "loss": 0.001, "step": 23632 }, { "epoch": 7.34, "learning_rate": 5.505101020204041e-07, "loss": 0.0019, "step": 23633 }, { "epoch": 7.34, "learning_rate": 5.501100220044009e-07, "loss": 0.0006, "step": 23634 }, { "epoch": 7.34, "learning_rate": 5.497099419883977e-07, "loss": 0.0006, "step": 23635 }, { "epoch": 7.34, "learning_rate": 5.493098619723945e-07, "loss": 0.0008, "step": 23636 }, { "epoch": 7.34, "learning_rate": 5.489097819563913e-07, "loss": 0.0013, "step": 23637 }, { "epoch": 7.34, "learning_rate": 5.485097019403881e-07, "loss": 0.0006, "step": 23638 }, { "epoch": 7.34, "learning_rate": 5.481096219243849e-07, "loss": 0.0011, "step": 23639 }, { "epoch": 7.34, "learning_rate": 5.477095419083817e-07, "loss": 0.0007, "step": 23640 }, { "epoch": 7.34, "learning_rate": 5.473094618923786e-07, "loss": 0.0007, "step": 23641 }, { "epoch": 7.34, "learning_rate": 5.469093818763753e-07, "loss": 0.001, "step": 23642 }, { "epoch": 7.34, "learning_rate": 5.465093018603721e-07, "loss": 0.0014, "step": 23643 }, { "epoch": 7.34, "learning_rate": 5.46109221844369e-07, "loss": 0.0174, "step": 23644 }, { "epoch": 7.34, "learning_rate": 5.457091418283658e-07, "loss": 0.0006, "step": 23645 }, { "epoch": 7.34, "learning_rate": 5.453090618123624e-07, "loss": 0.0008, "step": 23646 }, { "epoch": 7.34, "learning_rate": 5.449089817963593e-07, "loss": 0.0013, "step": 23647 }, { "epoch": 7.34, "learning_rate": 5.445089017803561e-07, "loss": 0.0013, "step": 23648 }, { "epoch": 7.34, "learning_rate": 5.441088217643529e-07, "loss": 0.0007, "step": 23649 }, { "epoch": 7.34, "learning_rate": 5.437087417483497e-07, "loss": 0.0007, "step": 23650 }, { "epoch": 7.34, "learning_rate": 5.433086617323465e-07, "loss": 0.0005, "step": 23651 }, { "epoch": 7.34, "learning_rate": 5.429085817163433e-07, "loss": 0.0012, "step": 23652 }, { "epoch": 7.34, "learning_rate": 5.425085017003401e-07, "loss": 0.001, "step": 23653 }, { "epoch": 7.34, "learning_rate": 5.421084216843369e-07, "loss": 0.0005, "step": 23654 }, { "epoch": 7.34, "learning_rate": 5.417083416683337e-07, "loss": 0.0011, "step": 23655 }, { "epoch": 7.34, "learning_rate": 5.413082616523306e-07, "loss": 0.0015, "step": 23656 }, { "epoch": 7.34, "learning_rate": 5.409081816363273e-07, "loss": 0.0008, "step": 23657 }, { "epoch": 7.34, "learning_rate": 5.405081016203241e-07, "loss": 0.0008, "step": 23658 }, { "epoch": 7.34, "learning_rate": 5.40108021604321e-07, "loss": 0.0007, "step": 23659 }, { "epoch": 7.34, "learning_rate": 5.397079415883177e-07, "loss": 0.0005, "step": 23660 }, { "epoch": 7.34, "learning_rate": 5.393078615723145e-07, "loss": 0.0035, "step": 23661 }, { "epoch": 7.34, "learning_rate": 5.389077815563113e-07, "loss": 0.0007, "step": 23662 }, { "epoch": 7.34, "learning_rate": 5.385077015403081e-07, "loss": 0.0009, "step": 23663 }, { "epoch": 7.34, "learning_rate": 5.381076215243049e-07, "loss": 0.0008, "step": 23664 }, { "epoch": 7.35, "learning_rate": 5.377075415083017e-07, "loss": 0.0036, "step": 23665 }, { "epoch": 7.35, "learning_rate": 5.373074614922985e-07, "loss": 0.0009, "step": 23666 }, { "epoch": 7.35, "learning_rate": 5.369073814762953e-07, "loss": 0.0016, "step": 23667 }, { "epoch": 7.35, "learning_rate": 5.365073014602921e-07, "loss": 0.0004, "step": 23668 }, { "epoch": 7.35, "learning_rate": 5.361072214442889e-07, "loss": 0.0007, "step": 23669 }, { "epoch": 7.35, "learning_rate": 5.357071414282857e-07, "loss": 0.0005, "step": 23670 }, { "epoch": 7.35, "learning_rate": 5.353070614122826e-07, "loss": 0.0011, "step": 23671 }, { "epoch": 7.35, "learning_rate": 5.349069813962793e-07, "loss": 0.0005, "step": 23672 }, { "epoch": 7.35, "learning_rate": 5.34506901380276e-07, "loss": 0.0006, "step": 23673 }, { "epoch": 7.35, "learning_rate": 5.341068213642729e-07, "loss": 0.0017, "step": 23674 }, { "epoch": 7.35, "learning_rate": 5.337067413482696e-07, "loss": 0.0006, "step": 23675 }, { "epoch": 7.35, "learning_rate": 5.333066613322665e-07, "loss": 0.0011, "step": 23676 }, { "epoch": 7.35, "learning_rate": 5.329065813162633e-07, "loss": 0.0006, "step": 23677 }, { "epoch": 7.35, "learning_rate": 5.325065013002601e-07, "loss": 0.0007, "step": 23678 }, { "epoch": 7.35, "learning_rate": 5.321064212842569e-07, "loss": 0.0013, "step": 23679 }, { "epoch": 7.35, "learning_rate": 5.317063412682537e-07, "loss": 0.001, "step": 23680 }, { "epoch": 7.35, "learning_rate": 5.313062612522505e-07, "loss": 0.0009, "step": 23681 }, { "epoch": 7.35, "learning_rate": 5.309061812362473e-07, "loss": 0.0007, "step": 23682 }, { "epoch": 7.35, "learning_rate": 5.305061012202441e-07, "loss": 0.0164, "step": 23683 }, { "epoch": 7.35, "learning_rate": 5.301060212042409e-07, "loss": 0.0261, "step": 23684 }, { "epoch": 7.35, "learning_rate": 5.297059411882377e-07, "loss": 0.001, "step": 23685 }, { "epoch": 7.35, "learning_rate": 5.293058611722346e-07, "loss": 0.0008, "step": 23686 }, { "epoch": 7.35, "learning_rate": 5.289057811562312e-07, "loss": 0.001, "step": 23687 }, { "epoch": 7.35, "learning_rate": 5.285057011402281e-07, "loss": 0.0003, "step": 23688 }, { "epoch": 7.35, "learning_rate": 5.281056211242249e-07, "loss": 0.0006, "step": 23689 }, { "epoch": 7.35, "learning_rate": 5.277055411082216e-07, "loss": 0.0009, "step": 23690 }, { "epoch": 7.35, "learning_rate": 5.273054610922185e-07, "loss": 0.0029, "step": 23691 }, { "epoch": 7.35, "learning_rate": 5.269053810762153e-07, "loss": 0.0008, "step": 23692 }, { "epoch": 7.35, "learning_rate": 5.265053010602121e-07, "loss": 0.0021, "step": 23693 }, { "epoch": 7.35, "learning_rate": 5.261052210442089e-07, "loss": 0.0012, "step": 23694 }, { "epoch": 7.35, "learning_rate": 5.257051410282057e-07, "loss": 0.0011, "step": 23695 }, { "epoch": 7.35, "learning_rate": 5.253050610122025e-07, "loss": 0.0019, "step": 23696 }, { "epoch": 7.36, "learning_rate": 5.249049809961993e-07, "loss": 0.0017, "step": 23697 }, { "epoch": 7.36, "learning_rate": 5.245049009801961e-07, "loss": 0.0009, "step": 23698 }, { "epoch": 7.36, "learning_rate": 5.241048209641929e-07, "loss": 0.001, "step": 23699 }, { "epoch": 7.36, "learning_rate": 5.237047409481896e-07, "loss": 0.0004, "step": 23700 }, { "epoch": 7.36, "learning_rate": 5.233046609321865e-07, "loss": 0.0004, "step": 23701 }, { "epoch": 7.36, "learning_rate": 5.229045809161832e-07, "loss": 0.0011, "step": 23702 }, { "epoch": 7.36, "learning_rate": 5.225045009001801e-07, "loss": 0.0005, "step": 23703 }, { "epoch": 7.36, "learning_rate": 5.221044208841769e-07, "loss": 0.0023, "step": 23704 }, { "epoch": 7.36, "learning_rate": 5.217043408681736e-07, "loss": 0.0016, "step": 23705 }, { "epoch": 7.36, "learning_rate": 5.213042608521705e-07, "loss": 0.001, "step": 23706 }, { "epoch": 7.36, "learning_rate": 5.209041808361673e-07, "loss": 0.0015, "step": 23707 }, { "epoch": 7.36, "learning_rate": 5.205041008201641e-07, "loss": 0.001, "step": 23708 }, { "epoch": 7.36, "learning_rate": 5.201040208041609e-07, "loss": 0.0008, "step": 23709 }, { "epoch": 7.36, "learning_rate": 5.197039407881577e-07, "loss": 0.0005, "step": 23710 }, { "epoch": 7.36, "learning_rate": 5.193038607721545e-07, "loss": 0.0006, "step": 23711 }, { "epoch": 7.36, "learning_rate": 5.189037807561513e-07, "loss": 0.0204, "step": 23712 }, { "epoch": 7.36, "learning_rate": 5.18503700740148e-07, "loss": 0.0009, "step": 23713 }, { "epoch": 7.36, "learning_rate": 5.181036207241448e-07, "loss": 0.0008, "step": 23714 }, { "epoch": 7.36, "learning_rate": 5.177035407081417e-07, "loss": 0.0013, "step": 23715 }, { "epoch": 7.36, "learning_rate": 5.173034606921384e-07, "loss": 0.001, "step": 23716 }, { "epoch": 7.36, "learning_rate": 5.169033806761352e-07, "loss": 0.0232, "step": 23717 }, { "epoch": 7.36, "learning_rate": 5.165033006601321e-07, "loss": 0.0011, "step": 23718 }, { "epoch": 7.36, "learning_rate": 5.161032206441289e-07, "loss": 0.0011, "step": 23719 }, { "epoch": 7.36, "learning_rate": 5.157031406281256e-07, "loss": 0.0003, "step": 23720 }, { "epoch": 7.36, "learning_rate": 5.153030606121225e-07, "loss": 0.0009, "step": 23721 }, { "epoch": 7.36, "learning_rate": 5.149029805961193e-07, "loss": 0.0005, "step": 23722 }, { "epoch": 7.36, "learning_rate": 5.145029005801161e-07, "loss": 0.0004, "step": 23723 }, { "epoch": 7.36, "learning_rate": 5.141028205641129e-07, "loss": 0.0005, "step": 23724 }, { "epoch": 7.36, "learning_rate": 5.137027405481097e-07, "loss": 0.0009, "step": 23725 }, { "epoch": 7.36, "learning_rate": 5.133026605321065e-07, "loss": 0.0005, "step": 23726 }, { "epoch": 7.36, "learning_rate": 5.129025805161032e-07, "loss": 0.0215, "step": 23727 }, { "epoch": 7.36, "learning_rate": 5.125025005001e-07, "loss": 0.0009, "step": 23728 }, { "epoch": 7.36, "learning_rate": 5.121024204840968e-07, "loss": 0.02, "step": 23729 }, { "epoch": 7.37, "learning_rate": 5.117023404680937e-07, "loss": 0.0017, "step": 23730 }, { "epoch": 7.37, "learning_rate": 5.113022604520904e-07, "loss": 0.0012, "step": 23731 }, { "epoch": 7.37, "learning_rate": 5.109021804360872e-07, "loss": 0.0005, "step": 23732 }, { "epoch": 7.37, "learning_rate": 5.105021004200841e-07, "loss": 0.0008, "step": 23733 }, { "epoch": 7.37, "learning_rate": 5.101020204040809e-07, "loss": 0.0006, "step": 23734 }, { "epoch": 7.37, "learning_rate": 5.097019403880776e-07, "loss": 0.0008, "step": 23735 }, { "epoch": 7.37, "learning_rate": 5.093018603720745e-07, "loss": 0.001, "step": 23736 }, { "epoch": 7.37, "learning_rate": 5.089017803560713e-07, "loss": 0.0014, "step": 23737 }, { "epoch": 7.37, "learning_rate": 5.085017003400681e-07, "loss": 0.0008, "step": 23738 }, { "epoch": 7.37, "learning_rate": 5.081016203240649e-07, "loss": 0.0083, "step": 23739 }, { "epoch": 7.37, "learning_rate": 5.077015403080616e-07, "loss": 0.0005, "step": 23740 }, { "epoch": 7.37, "learning_rate": 5.073014602920584e-07, "loss": 0.0008, "step": 23741 }, { "epoch": 7.37, "learning_rate": 5.069013802760552e-07, "loss": 0.0008, "step": 23742 }, { "epoch": 7.37, "learning_rate": 5.06501300260052e-07, "loss": 0.0006, "step": 23743 }, { "epoch": 7.37, "learning_rate": 5.061012202440488e-07, "loss": 0.0014, "step": 23744 }, { "epoch": 7.37, "learning_rate": 5.057011402280457e-07, "loss": 0.0005, "step": 23745 }, { "epoch": 7.37, "learning_rate": 5.053010602120424e-07, "loss": 0.0012, "step": 23746 }, { "epoch": 7.37, "learning_rate": 5.049009801960392e-07, "loss": 0.0007, "step": 23747 }, { "epoch": 7.37, "learning_rate": 5.045009001800361e-07, "loss": 0.0012, "step": 23748 }, { "epoch": 7.37, "learning_rate": 5.041008201640329e-07, "loss": 0.0018, "step": 23749 }, { "epoch": 7.37, "learning_rate": 5.037007401480297e-07, "loss": 0.0007, "step": 23750 }, { "epoch": 7.37, "learning_rate": 5.033006601320265e-07, "loss": 0.0019, "step": 23751 }, { "epoch": 7.37, "learning_rate": 5.029005801160233e-07, "loss": 0.0008, "step": 23752 }, { "epoch": 7.37, "learning_rate": 5.0250050010002e-07, "loss": 0.0003, "step": 23753 }, { "epoch": 7.37, "learning_rate": 5.021004200840168e-07, "loss": 0.0005, "step": 23754 }, { "epoch": 7.37, "learning_rate": 5.017003400680136e-07, "loss": 0.0006, "step": 23755 }, { "epoch": 7.37, "learning_rate": 5.013002600520104e-07, "loss": 0.0061, "step": 23756 }, { "epoch": 7.37, "learning_rate": 5.009001800360072e-07, "loss": 0.0011, "step": 23757 }, { "epoch": 7.37, "learning_rate": 5.00500100020004e-07, "loss": 0.0008, "step": 23758 }, { "epoch": 7.37, "learning_rate": 5.001000200040008e-07, "loss": 0.0008, "step": 23759 }, { "epoch": 7.37, "learning_rate": 4.996999399879977e-07, "loss": 0.0009, "step": 23760 }, { "epoch": 7.37, "learning_rate": 4.992998599719944e-07, "loss": 0.0015, "step": 23761 }, { "epoch": 7.38, "learning_rate": 4.988997799559912e-07, "loss": 0.001, "step": 23762 }, { "epoch": 7.38, "learning_rate": 4.984996999399881e-07, "loss": 0.0019, "step": 23763 }, { "epoch": 7.38, "learning_rate": 4.980996199239849e-07, "loss": 0.0056, "step": 23764 }, { "epoch": 7.38, "learning_rate": 4.976995399079817e-07, "loss": 0.0008, "step": 23765 }, { "epoch": 7.38, "learning_rate": 4.972994598919785e-07, "loss": 0.0007, "step": 23766 }, { "epoch": 7.38, "learning_rate": 4.968993798759752e-07, "loss": 0.0088, "step": 23767 }, { "epoch": 7.38, "learning_rate": 4.96499299859972e-07, "loss": 0.0005, "step": 23768 }, { "epoch": 7.38, "learning_rate": 4.960992198439688e-07, "loss": 0.0011, "step": 23769 }, { "epoch": 7.38, "learning_rate": 4.956991398279656e-07, "loss": 0.0004, "step": 23770 }, { "epoch": 7.38, "learning_rate": 4.952990598119624e-07, "loss": 0.0006, "step": 23771 }, { "epoch": 7.38, "learning_rate": 4.948989797959592e-07, "loss": 0.0182, "step": 23772 }, { "epoch": 7.38, "learning_rate": 4.94498899779956e-07, "loss": 0.0005, "step": 23773 }, { "epoch": 7.38, "learning_rate": 4.940988197639528e-07, "loss": 0.0008, "step": 23774 }, { "epoch": 7.38, "learning_rate": 4.936987397479497e-07, "loss": 0.0013, "step": 23775 }, { "epoch": 7.38, "learning_rate": 4.932986597319464e-07, "loss": 0.0005, "step": 23776 }, { "epoch": 7.38, "learning_rate": 4.928985797159432e-07, "loss": 0.001, "step": 23777 }, { "epoch": 7.38, "learning_rate": 4.928985797159432e-07, "loss": 0.0327, "step": 23778 }, { "epoch": 7.38, "learning_rate": 4.924984996999401e-07, "loss": 0.0011, "step": 23779 }, { "epoch": 7.38, "learning_rate": 4.920984196839369e-07, "loss": 0.0022, "step": 23780 }, { "epoch": 7.38, "learning_rate": 4.916983396679337e-07, "loss": 0.0038, "step": 23781 }, { "epoch": 7.38, "learning_rate": 4.912982596519304e-07, "loss": 0.0006, "step": 23782 }, { "epoch": 7.38, "learning_rate": 4.908981796359272e-07, "loss": 0.0012, "step": 23783 }, { "epoch": 7.38, "learning_rate": 4.90498099619924e-07, "loss": 0.0007, "step": 23784 }, { "epoch": 7.38, "learning_rate": 4.900980196039208e-07, "loss": 0.0009, "step": 23785 }, { "epoch": 7.38, "learning_rate": 4.896979395879176e-07, "loss": 0.0012, "step": 23786 }, { "epoch": 7.38, "learning_rate": 4.892978595719144e-07, "loss": 0.0008, "step": 23787 }, { "epoch": 7.38, "learning_rate": 4.888977795559112e-07, "loss": 0.0004, "step": 23788 }, { "epoch": 7.38, "learning_rate": 4.88497699539908e-07, "loss": 0.0005, "step": 23789 }, { "epoch": 7.38, "learning_rate": 4.880976195239048e-07, "loss": 0.0008, "step": 23790 }, { "epoch": 7.38, "learning_rate": 4.876975395079017e-07, "loss": 0.0007, "step": 23791 }, { "epoch": 7.38, "learning_rate": 4.872974594918984e-07, "loss": 0.0005, "step": 23792 }, { "epoch": 7.38, "learning_rate": 4.868973794758953e-07, "loss": 0.0009, "step": 23793 }, { "epoch": 7.39, "learning_rate": 4.864972994598921e-07, "loss": 0.0006, "step": 23794 }, { "epoch": 7.39, "learning_rate": 4.860972194438888e-07, "loss": 0.0012, "step": 23795 }, { "epoch": 7.39, "learning_rate": 4.856971394278856e-07, "loss": 0.0011, "step": 23796 }, { "epoch": 7.39, "learning_rate": 4.852970594118824e-07, "loss": 0.0005, "step": 23797 }, { "epoch": 7.39, "learning_rate": 4.848969793958792e-07, "loss": 0.0007, "step": 23798 }, { "epoch": 7.39, "learning_rate": 4.84496899379876e-07, "loss": 0.0008, "step": 23799 }, { "epoch": 7.39, "learning_rate": 4.840968193638728e-07, "loss": 0.0007, "step": 23800 }, { "epoch": 7.39, "learning_rate": 4.836967393478696e-07, "loss": 0.0016, "step": 23801 }, { "epoch": 7.39, "learning_rate": 4.832966593318664e-07, "loss": 0.001, "step": 23802 }, { "epoch": 7.39, "learning_rate": 4.828965793158632e-07, "loss": 0.0009, "step": 23803 }, { "epoch": 7.39, "learning_rate": 4.8249649929986e-07, "loss": 0.001, "step": 23804 }, { "epoch": 7.39, "learning_rate": 4.820964192838568e-07, "loss": 0.001, "step": 23805 }, { "epoch": 7.39, "learning_rate": 4.816963392678537e-07, "loss": 0.0066, "step": 23806 }, { "epoch": 7.39, "learning_rate": 4.812962592518504e-07, "loss": 0.0006, "step": 23807 }, { "epoch": 7.39, "learning_rate": 4.808961792358473e-07, "loss": 0.0005, "step": 23808 }, { "epoch": 7.39, "learning_rate": 4.80496099219844e-07, "loss": 0.0013, "step": 23809 }, { "epoch": 7.39, "learning_rate": 4.800960192038407e-07, "loss": 0.0006, "step": 23810 }, { "epoch": 7.39, "learning_rate": 4.796959391878376e-07, "loss": 0.0006, "step": 23811 }, { "epoch": 7.39, "learning_rate": 4.792958591718344e-07, "loss": 0.001, "step": 23812 }, { "epoch": 7.39, "learning_rate": 4.788957791558312e-07, "loss": 0.0007, "step": 23813 }, { "epoch": 7.39, "learning_rate": 4.78495699139828e-07, "loss": 0.0004, "step": 23814 }, { "epoch": 7.39, "learning_rate": 4.780956191238248e-07, "loss": 0.0007, "step": 23815 }, { "epoch": 7.39, "learning_rate": 4.776955391078216e-07, "loss": 0.0007, "step": 23816 }, { "epoch": 7.39, "learning_rate": 4.772954590918184e-07, "loss": 0.0008, "step": 23817 }, { "epoch": 7.39, "learning_rate": 4.768953790758152e-07, "loss": 0.0006, "step": 23818 }, { "epoch": 7.39, "learning_rate": 4.76495299059812e-07, "loss": 0.001, "step": 23819 }, { "epoch": 7.39, "learning_rate": 4.760952190438088e-07, "loss": 0.0006, "step": 23820 }, { "epoch": 7.39, "learning_rate": 4.7569513902780566e-07, "loss": 0.0004, "step": 23821 }, { "epoch": 7.39, "learning_rate": 4.752950590118024e-07, "loss": 0.0043, "step": 23822 }, { "epoch": 7.39, "learning_rate": 4.748949789957992e-07, "loss": 0.0011, "step": 23823 }, { "epoch": 7.39, "learning_rate": 4.7449489897979603e-07, "loss": 0.0005, "step": 23824 }, { "epoch": 7.39, "learning_rate": 4.7409481896379277e-07, "loss": 0.0007, "step": 23825 }, { "epoch": 7.4, "learning_rate": 4.7369473894778956e-07, "loss": 0.0009, "step": 23826 }, { "epoch": 7.4, "learning_rate": 4.732946589317864e-07, "loss": 0.001, "step": 23827 }, { "epoch": 7.4, "learning_rate": 4.728945789157832e-07, "loss": 0.0005, "step": 23828 }, { "epoch": 7.4, "learning_rate": 4.7249449889978e-07, "loss": 0.0011, "step": 23829 }, { "epoch": 7.4, "learning_rate": 4.720944188837768e-07, "loss": 0.0019, "step": 23830 }, { "epoch": 7.4, "learning_rate": 4.716943388677736e-07, "loss": 0.001, "step": 23831 }, { "epoch": 7.4, "learning_rate": 4.712942588517704e-07, "loss": 0.0013, "step": 23832 }, { "epoch": 7.4, "learning_rate": 4.7089417883576716e-07, "loss": 0.0011, "step": 23833 }, { "epoch": 7.4, "learning_rate": 4.70494098819764e-07, "loss": 0.0008, "step": 23834 }, { "epoch": 7.4, "learning_rate": 4.700940188037608e-07, "loss": 0.0013, "step": 23835 }, { "epoch": 7.4, "learning_rate": 4.6969393878775754e-07, "loss": 0.0005, "step": 23836 }, { "epoch": 7.4, "learning_rate": 4.692938587717544e-07, "loss": 0.0008, "step": 23837 }, { "epoch": 7.4, "learning_rate": 4.688937787557512e-07, "loss": 0.0011, "step": 23838 }, { "epoch": 7.4, "learning_rate": 4.68493698739748e-07, "loss": 0.0011, "step": 23839 }, { "epoch": 7.4, "learning_rate": 4.6809361872374476e-07, "loss": 0.0009, "step": 23840 }, { "epoch": 7.4, "learning_rate": 4.6769353870774155e-07, "loss": 0.0009, "step": 23841 }, { "epoch": 7.4, "learning_rate": 4.672934586917384e-07, "loss": 0.0008, "step": 23842 }, { "epoch": 7.4, "learning_rate": 4.668933786757352e-07, "loss": 0.0005, "step": 23843 }, { "epoch": 7.4, "learning_rate": 4.66493298659732e-07, "loss": 0.0007, "step": 23844 }, { "epoch": 7.4, "learning_rate": 4.660932186437288e-07, "loss": 0.0007, "step": 23845 }, { "epoch": 7.4, "learning_rate": 4.656931386277256e-07, "loss": 0.0005, "step": 23846 }, { "epoch": 7.4, "learning_rate": 4.652930586117224e-07, "loss": 0.0005, "step": 23847 }, { "epoch": 7.4, "learning_rate": 4.6489297859571915e-07, "loss": 0.0004, "step": 23848 }, { "epoch": 7.4, "learning_rate": 4.64492898579716e-07, "loss": 0.0025, "step": 23849 }, { "epoch": 7.4, "learning_rate": 4.640928185637128e-07, "loss": 0.0014, "step": 23850 }, { "epoch": 7.4, "learning_rate": 4.6369273854770953e-07, "loss": 0.0029, "step": 23851 }, { "epoch": 7.4, "learning_rate": 4.6329265853170637e-07, "loss": 0.001, "step": 23852 }, { "epoch": 7.4, "learning_rate": 4.6289257851570316e-07, "loss": 0.0037, "step": 23853 }, { "epoch": 7.4, "learning_rate": 4.624924984997e-07, "loss": 0.0233, "step": 23854 }, { "epoch": 7.4, "learning_rate": 4.6209241848369675e-07, "loss": 0.0003, "step": 23855 }, { "epoch": 7.4, "learning_rate": 4.616923384676936e-07, "loss": 0.0012, "step": 23856 }, { "epoch": 7.4, "learning_rate": 4.612922584516904e-07, "loss": 0.001, "step": 23857 }, { "epoch": 7.41, "learning_rate": 4.608921784356872e-07, "loss": 0.0007, "step": 23858 }, { "epoch": 7.41, "learning_rate": 4.6049209841968397e-07, "loss": 0.0009, "step": 23859 }, { "epoch": 7.41, "learning_rate": 4.6009201840368076e-07, "loss": 0.0013, "step": 23860 }, { "epoch": 7.41, "learning_rate": 4.596919383876776e-07, "loss": 0.0005, "step": 23861 }, { "epoch": 7.41, "learning_rate": 4.592918583716744e-07, "loss": 0.0009, "step": 23862 }, { "epoch": 7.41, "learning_rate": 4.5889177835567114e-07, "loss": 0.0003, "step": 23863 }, { "epoch": 7.41, "learning_rate": 4.58491698339668e-07, "loss": 0.0019, "step": 23864 }, { "epoch": 7.41, "learning_rate": 4.580916183236648e-07, "loss": 0.0219, "step": 23865 }, { "epoch": 7.41, "learning_rate": 4.576915383076615e-07, "loss": 0.0004, "step": 23866 }, { "epoch": 7.41, "learning_rate": 4.5729145829165836e-07, "loss": 0.0162, "step": 23867 }, { "epoch": 7.41, "learning_rate": 4.5689137827565515e-07, "loss": 0.0006, "step": 23868 }, { "epoch": 7.41, "learning_rate": 4.56491298259652e-07, "loss": 0.0005, "step": 23869 }, { "epoch": 7.41, "learning_rate": 4.5609121824364874e-07, "loss": 0.0011, "step": 23870 }, { "epoch": 7.41, "learning_rate": 4.556911382276456e-07, "loss": 0.0185, "step": 23871 }, { "epoch": 7.41, "learning_rate": 4.552910582116424e-07, "loss": 0.0283, "step": 23872 }, { "epoch": 7.41, "learning_rate": 4.548909781956392e-07, "loss": 0.0051, "step": 23873 }, { "epoch": 7.41, "learning_rate": 4.5449089817963596e-07, "loss": 0.0009, "step": 23874 }, { "epoch": 7.41, "learning_rate": 4.5409081816363275e-07, "loss": 0.0008, "step": 23875 }, { "epoch": 7.41, "learning_rate": 4.536907381476296e-07, "loss": 0.0011, "step": 23876 }, { "epoch": 7.41, "learning_rate": 4.5329065813162634e-07, "loss": 0.0015, "step": 23877 }, { "epoch": 7.41, "learning_rate": 4.5289057811562313e-07, "loss": 0.0004, "step": 23878 }, { "epoch": 7.41, "learning_rate": 4.5249049809961997e-07, "loss": 0.0006, "step": 23879 }, { "epoch": 7.41, "learning_rate": 4.5209041808361676e-07, "loss": 0.0007, "step": 23880 }, { "epoch": 7.41, "learning_rate": 4.5169033806761356e-07, "loss": 0.0005, "step": 23881 }, { "epoch": 7.41, "learning_rate": 4.5129025805161035e-07, "loss": 0.0006, "step": 23882 }, { "epoch": 7.41, "learning_rate": 4.508901780356072e-07, "loss": 0.0009, "step": 23883 }, { "epoch": 7.41, "learning_rate": 4.50490098019604e-07, "loss": 0.0003, "step": 23884 }, { "epoch": 7.41, "learning_rate": 4.500900180036007e-07, "loss": 0.0011, "step": 23885 }, { "epoch": 7.41, "learning_rate": 4.4968993798759757e-07, "loss": 0.0212, "step": 23886 }, { "epoch": 7.41, "learning_rate": 4.4928985797159436e-07, "loss": 0.0009, "step": 23887 }, { "epoch": 7.41, "learning_rate": 4.488897779555912e-07, "loss": 0.0004, "step": 23888 }, { "epoch": 7.41, "learning_rate": 4.4848969793958795e-07, "loss": 0.0042, "step": 23889 }, { "epoch": 7.41, "learning_rate": 4.4808961792358474e-07, "loss": 0.0009, "step": 23890 }, { "epoch": 7.42, "learning_rate": 4.476895379075816e-07, "loss": 0.0006, "step": 23891 }, { "epoch": 7.42, "learning_rate": 4.472894578915783e-07, "loss": 0.0005, "step": 23892 }, { "epoch": 7.42, "learning_rate": 4.468893778755751e-07, "loss": 0.001, "step": 23893 }, { "epoch": 7.42, "learning_rate": 4.4648929785957196e-07, "loss": 0.0012, "step": 23894 }, { "epoch": 7.42, "learning_rate": 4.4608921784356875e-07, "loss": 0.0036, "step": 23895 }, { "epoch": 7.42, "learning_rate": 4.4568913782756554e-07, "loss": 0.0031, "step": 23896 }, { "epoch": 7.42, "learning_rate": 4.4528905781156234e-07, "loss": 0.0009, "step": 23897 }, { "epoch": 7.42, "learning_rate": 4.448889777955592e-07, "loss": 0.0011, "step": 23898 }, { "epoch": 7.42, "learning_rate": 4.44488897779556e-07, "loss": 0.0023, "step": 23899 }, { "epoch": 7.42, "learning_rate": 4.440888177635527e-07, "loss": 0.0014, "step": 23900 }, { "epoch": 7.42, "learning_rate": 4.4368873774754956e-07, "loss": 0.0004, "step": 23901 }, { "epoch": 7.42, "learning_rate": 4.4328865773154635e-07, "loss": 0.0007, "step": 23902 }, { "epoch": 7.42, "learning_rate": 4.428885777155431e-07, "loss": 0.0026, "step": 23903 }, { "epoch": 7.42, "learning_rate": 4.4248849769953993e-07, "loss": 0.0013, "step": 23904 }, { "epoch": 7.42, "learning_rate": 4.4208841768353673e-07, "loss": 0.0008, "step": 23905 }, { "epoch": 7.42, "learning_rate": 4.4168833766753357e-07, "loss": 0.0012, "step": 23906 }, { "epoch": 7.42, "learning_rate": 4.412882576515303e-07, "loss": 0.0008, "step": 23907 }, { "epoch": 7.42, "learning_rate": 4.4088817763552716e-07, "loss": 0.0004, "step": 23908 }, { "epoch": 7.42, "learning_rate": 4.4048809761952395e-07, "loss": 0.0008, "step": 23909 }, { "epoch": 7.42, "learning_rate": 4.400880176035208e-07, "loss": 0.001, "step": 23910 }, { "epoch": 7.42, "learning_rate": 4.3968793758751753e-07, "loss": 0.0602, "step": 23911 }, { "epoch": 7.42, "learning_rate": 4.392878575715143e-07, "loss": 0.0008, "step": 23912 }, { "epoch": 7.42, "learning_rate": 4.3888777755551117e-07, "loss": 0.0008, "step": 23913 }, { "epoch": 7.42, "learning_rate": 4.3848769753950796e-07, "loss": 0.0021, "step": 23914 }, { "epoch": 7.42, "learning_rate": 4.380876175235047e-07, "loss": 0.0007, "step": 23915 }, { "epoch": 7.42, "learning_rate": 4.3768753750750155e-07, "loss": 0.0007, "step": 23916 }, { "epoch": 7.42, "learning_rate": 4.3728745749149834e-07, "loss": 0.0013, "step": 23917 }, { "epoch": 7.42, "learning_rate": 4.368873774754951e-07, "loss": 0.0005, "step": 23918 }, { "epoch": 7.42, "learning_rate": 4.364872974594919e-07, "loss": 0.0004, "step": 23919 }, { "epoch": 7.42, "learning_rate": 4.360872174434887e-07, "loss": 0.0067, "step": 23920 }, { "epoch": 7.42, "learning_rate": 4.3568713742748556e-07, "loss": 0.0008, "step": 23921 }, { "epoch": 7.42, "learning_rate": 4.352870574114823e-07, "loss": 0.0008, "step": 23922 }, { "epoch": 7.43, "learning_rate": 4.3488697739547914e-07, "loss": 0.0011, "step": 23923 }, { "epoch": 7.43, "learning_rate": 4.3448689737947594e-07, "loss": 0.0004, "step": 23924 }, { "epoch": 7.43, "learning_rate": 4.340868173634728e-07, "loss": 0.0007, "step": 23925 }, { "epoch": 7.43, "learning_rate": 4.336867373474695e-07, "loss": 0.0011, "step": 23926 }, { "epoch": 7.43, "learning_rate": 4.332866573314663e-07, "loss": 0.0006, "step": 23927 }, { "epoch": 7.43, "learning_rate": 4.3288657731546316e-07, "loss": 0.0042, "step": 23928 }, { "epoch": 7.43, "learning_rate": 4.3248649729945995e-07, "loss": 0.0006, "step": 23929 }, { "epoch": 7.43, "learning_rate": 4.320864172834567e-07, "loss": 0.001, "step": 23930 }, { "epoch": 7.43, "learning_rate": 4.3168633726745353e-07, "loss": 0.0008, "step": 23931 }, { "epoch": 7.43, "learning_rate": 4.3128625725145033e-07, "loss": 0.0027, "step": 23932 }, { "epoch": 7.43, "learning_rate": 4.308861772354471e-07, "loss": 0.001, "step": 23933 }, { "epoch": 7.43, "learning_rate": 4.304860972194439e-07, "loss": 0.001, "step": 23934 }, { "epoch": 7.43, "learning_rate": 4.3008601720344076e-07, "loss": 0.0005, "step": 23935 }, { "epoch": 7.43, "learning_rate": 4.2968593718743755e-07, "loss": 0.0013, "step": 23936 }, { "epoch": 7.43, "learning_rate": 4.292858571714343e-07, "loss": 0.0062, "step": 23937 }, { "epoch": 7.43, "learning_rate": 4.2888577715543113e-07, "loss": 0.0011, "step": 23938 }, { "epoch": 7.43, "learning_rate": 4.284856971394279e-07, "loss": 0.0119, "step": 23939 }, { "epoch": 7.43, "learning_rate": 4.2808561712342477e-07, "loss": 0.0008, "step": 23940 }, { "epoch": 7.43, "learning_rate": 4.276855371074215e-07, "loss": 0.0005, "step": 23941 }, { "epoch": 7.43, "learning_rate": 4.272854570914183e-07, "loss": 0.0011, "step": 23942 }, { "epoch": 7.43, "learning_rate": 4.2688537707541515e-07, "loss": 0.0006, "step": 23943 }, { "epoch": 7.43, "learning_rate": 4.264852970594119e-07, "loss": 0.0005, "step": 23944 }, { "epoch": 7.43, "learning_rate": 4.260852170434087e-07, "loss": 0.0009, "step": 23945 }, { "epoch": 7.43, "learning_rate": 4.256851370274055e-07, "loss": 0.0009, "step": 23946 }, { "epoch": 7.43, "learning_rate": 4.252850570114023e-07, "loss": 0.0012, "step": 23947 }, { "epoch": 7.43, "learning_rate": 4.248849769953991e-07, "loss": 0.0007, "step": 23948 }, { "epoch": 7.43, "learning_rate": 4.244848969793959e-07, "loss": 0.0003, "step": 23949 }, { "epoch": 7.43, "learning_rate": 4.2408481696339274e-07, "loss": 0.0235, "step": 23950 }, { "epoch": 7.43, "learning_rate": 4.2368473694738954e-07, "loss": 0.0007, "step": 23951 }, { "epoch": 7.43, "learning_rate": 4.232846569313863e-07, "loss": 0.0014, "step": 23952 }, { "epoch": 7.43, "learning_rate": 4.228845769153831e-07, "loss": 0.0006, "step": 23953 }, { "epoch": 7.43, "learning_rate": 4.224844968993799e-07, "loss": 0.0013, "step": 23954 }, { "epoch": 7.44, "learning_rate": 4.2208441688337676e-07, "loss": 0.0006, "step": 23955 }, { "epoch": 7.44, "learning_rate": 4.216843368673735e-07, "loss": 0.0007, "step": 23956 }, { "epoch": 7.44, "learning_rate": 4.212842568513703e-07, "loss": 0.0005, "step": 23957 }, { "epoch": 7.44, "learning_rate": 4.2088417683536713e-07, "loss": 0.0006, "step": 23958 }, { "epoch": 7.44, "learning_rate": 4.204840968193639e-07, "loss": 0.0006, "step": 23959 }, { "epoch": 7.44, "learning_rate": 4.200840168033607e-07, "loss": 0.0007, "step": 23960 }, { "epoch": 7.44, "learning_rate": 4.196839367873575e-07, "loss": 0.0008, "step": 23961 }, { "epoch": 7.44, "learning_rate": 4.1928385677135436e-07, "loss": 0.0007, "step": 23962 }, { "epoch": 7.44, "learning_rate": 4.188837767553511e-07, "loss": 0.0009, "step": 23963 }, { "epoch": 7.44, "learning_rate": 4.184836967393479e-07, "loss": 0.0011, "step": 23964 }, { "epoch": 7.44, "learning_rate": 4.1808361672334473e-07, "loss": 0.0045, "step": 23965 }, { "epoch": 7.44, "learning_rate": 4.176835367073415e-07, "loss": 0.0026, "step": 23966 }, { "epoch": 7.44, "learning_rate": 4.1728345669133826e-07, "loss": 0.0011, "step": 23967 }, { "epoch": 7.44, "learning_rate": 4.168833766753351e-07, "loss": 0.0028, "step": 23968 }, { "epoch": 7.44, "learning_rate": 4.164832966593319e-07, "loss": 0.0205, "step": 23969 }, { "epoch": 7.44, "learning_rate": 4.1608321664332864e-07, "loss": 0.0005, "step": 23970 }, { "epoch": 7.44, "learning_rate": 4.156831366273255e-07, "loss": 0.0005, "step": 23971 }, { "epoch": 7.44, "learning_rate": 4.152830566113223e-07, "loss": 0.0209, "step": 23972 }, { "epoch": 7.44, "learning_rate": 4.148829765953191e-07, "loss": 0.0005, "step": 23973 }, { "epoch": 7.44, "learning_rate": 4.1448289657931586e-07, "loss": 0.0008, "step": 23974 }, { "epoch": 7.44, "learning_rate": 4.140828165633127e-07, "loss": 0.0007, "step": 23975 }, { "epoch": 7.44, "learning_rate": 4.136827365473095e-07, "loss": 0.0005, "step": 23976 }, { "epoch": 7.44, "learning_rate": 4.1328265653130634e-07, "loss": 0.0006, "step": 23977 }, { "epoch": 7.44, "learning_rate": 4.128825765153031e-07, "loss": 0.0012, "step": 23978 }, { "epoch": 7.44, "learning_rate": 4.124824964992999e-07, "loss": 0.0012, "step": 23979 }, { "epoch": 7.44, "learning_rate": 4.120824164832967e-07, "loss": 0.001, "step": 23980 }, { "epoch": 7.44, "learning_rate": 4.116823364672935e-07, "loss": 0.001, "step": 23981 }, { "epoch": 7.44, "learning_rate": 4.1128225645129025e-07, "loss": 0.0076, "step": 23982 }, { "epoch": 7.44, "learning_rate": 4.108821764352871e-07, "loss": 0.0563, "step": 23983 }, { "epoch": 7.44, "learning_rate": 4.104820964192839e-07, "loss": 0.001, "step": 23984 }, { "epoch": 7.44, "learning_rate": 4.100820164032807e-07, "loss": 0.0009, "step": 23985 }, { "epoch": 7.44, "learning_rate": 4.096819363872775e-07, "loss": 0.0008, "step": 23986 }, { "epoch": 7.45, "learning_rate": 4.092818563712743e-07, "loss": 0.0007, "step": 23987 }, { "epoch": 7.45, "learning_rate": 4.088817763552711e-07, "loss": 0.0018, "step": 23988 }, { "epoch": 7.45, "learning_rate": 4.0848169633926785e-07, "loss": 0.0007, "step": 23989 }, { "epoch": 7.45, "learning_rate": 4.080816163232647e-07, "loss": 0.0018, "step": 23990 }, { "epoch": 7.45, "learning_rate": 4.076815363072615e-07, "loss": 0.0229, "step": 23991 }, { "epoch": 7.45, "learning_rate": 4.0728145629125833e-07, "loss": 0.0006, "step": 23992 }, { "epoch": 7.45, "learning_rate": 4.0688137627525507e-07, "loss": 0.0049, "step": 23993 }, { "epoch": 7.45, "learning_rate": 4.0648129625925186e-07, "loss": 0.0006, "step": 23994 }, { "epoch": 7.45, "learning_rate": 4.060812162432487e-07, "loss": 0.0008, "step": 23995 }, { "epoch": 7.45, "learning_rate": 4.0568113622724545e-07, "loss": 0.0006, "step": 23996 }, { "epoch": 7.45, "learning_rate": 4.0528105621124224e-07, "loss": 0.0008, "step": 23997 }, { "epoch": 7.45, "learning_rate": 4.048809761952391e-07, "loss": 0.001, "step": 23998 }, { "epoch": 7.45, "learning_rate": 4.044808961792359e-07, "loss": 0.0014, "step": 23999 }, { "epoch": 7.45, "learning_rate": 4.0408081616323267e-07, "loss": 0.0006, "step": 24000 }, { "epoch": 7.45, "learning_rate": 4.0368073614722946e-07, "loss": 0.0017, "step": 24001 }, { "epoch": 7.45, "learning_rate": 4.032806561312263e-07, "loss": 0.0007, "step": 24002 }, { "epoch": 7.45, "learning_rate": 4.028805761152231e-07, "loss": 0.0015, "step": 24003 }, { "epoch": 7.45, "learning_rate": 4.0248049609921984e-07, "loss": 0.0009, "step": 24004 }, { "epoch": 7.45, "learning_rate": 4.020804160832167e-07, "loss": 0.0007, "step": 24005 }, { "epoch": 7.45, "learning_rate": 4.016803360672135e-07, "loss": 0.0006, "step": 24006 }, { "epoch": 7.45, "learning_rate": 4.012802560512103e-07, "loss": 0.0004, "step": 24007 }, { "epoch": 7.45, "learning_rate": 4.0088017603520706e-07, "loss": 0.0007, "step": 24008 }, { "epoch": 7.45, "learning_rate": 4.0048009601920385e-07, "loss": 0.0006, "step": 24009 }, { "epoch": 7.45, "learning_rate": 4.000800160032007e-07, "loss": 0.0008, "step": 24010 }, { "epoch": 7.45, "learning_rate": 3.9967993598719744e-07, "loss": 0.0006, "step": 24011 }, { "epoch": 7.45, "learning_rate": 3.992798559711943e-07, "loss": 0.0036, "step": 24012 }, { "epoch": 7.45, "learning_rate": 3.988797759551911e-07, "loss": 0.0009, "step": 24013 }, { "epoch": 7.45, "learning_rate": 3.984796959391879e-07, "loss": 0.0006, "step": 24014 }, { "epoch": 7.45, "learning_rate": 3.9807961592318466e-07, "loss": 0.0013, "step": 24015 }, { "epoch": 7.45, "learning_rate": 3.9767953590718145e-07, "loss": 0.01, "step": 24016 }, { "epoch": 7.45, "learning_rate": 3.972794558911783e-07, "loss": 0.0009, "step": 24017 }, { "epoch": 7.45, "learning_rate": 3.968793758751751e-07, "loss": 0.0007, "step": 24018 }, { "epoch": 7.45, "learning_rate": 3.9647929585917183e-07, "loss": 0.0012, "step": 24019 }, { "epoch": 7.46, "learning_rate": 3.9607921584316867e-07, "loss": 0.0006, "step": 24020 }, { "epoch": 7.46, "learning_rate": 3.9567913582716546e-07, "loss": 0.0008, "step": 24021 }, { "epoch": 7.46, "learning_rate": 3.952790558111623e-07, "loss": 0.0006, "step": 24022 }, { "epoch": 7.46, "learning_rate": 3.9487897579515905e-07, "loss": 0.0011, "step": 24023 }, { "epoch": 7.46, "learning_rate": 3.9447889577915584e-07, "loss": 0.0016, "step": 24024 }, { "epoch": 7.46, "learning_rate": 3.940788157631527e-07, "loss": 0.0046, "step": 24025 }, { "epoch": 7.46, "learning_rate": 3.936787357471494e-07, "loss": 0.0008, "step": 24026 }, { "epoch": 7.46, "learning_rate": 3.9327865573114627e-07, "loss": 0.0008, "step": 24027 }, { "epoch": 7.46, "learning_rate": 3.9287857571514306e-07, "loss": 0.0008, "step": 24028 }, { "epoch": 7.46, "learning_rate": 3.924784956991399e-07, "loss": 0.0005, "step": 24029 }, { "epoch": 7.46, "learning_rate": 3.9207841568313665e-07, "loss": 0.0007, "step": 24030 }, { "epoch": 7.46, "learning_rate": 3.9167833566713344e-07, "loss": 0.0012, "step": 24031 }, { "epoch": 7.46, "learning_rate": 3.912782556511303e-07, "loss": 0.0006, "step": 24032 }, { "epoch": 7.46, "learning_rate": 3.908781756351271e-07, "loss": 0.0014, "step": 24033 }, { "epoch": 7.46, "learning_rate": 3.904780956191238e-07, "loss": 0.0012, "step": 24034 }, { "epoch": 7.46, "learning_rate": 3.9007801560312066e-07, "loss": 0.0006, "step": 24035 }, { "epoch": 7.46, "learning_rate": 3.8967793558711745e-07, "loss": 0.0027, "step": 24036 }, { "epoch": 7.46, "learning_rate": 3.8927785557111424e-07, "loss": 0.0134, "step": 24037 }, { "epoch": 7.46, "learning_rate": 3.8887777555511104e-07, "loss": 0.0004, "step": 24038 }, { "epoch": 7.46, "learning_rate": 3.884776955391079e-07, "loss": 0.0007, "step": 24039 }, { "epoch": 7.46, "learning_rate": 3.880776155231047e-07, "loss": 0.0049, "step": 24040 }, { "epoch": 7.46, "learning_rate": 3.876775355071014e-07, "loss": 0.0037, "step": 24041 }, { "epoch": 7.46, "learning_rate": 3.8727745549109826e-07, "loss": 0.0048, "step": 24042 }, { "epoch": 7.46, "learning_rate": 3.8687737547509505e-07, "loss": 0.0025, "step": 24043 }, { "epoch": 7.46, "learning_rate": 3.864772954590919e-07, "loss": 0.0008, "step": 24044 }, { "epoch": 7.46, "learning_rate": 3.8607721544308863e-07, "loss": 0.0005, "step": 24045 }, { "epoch": 7.46, "learning_rate": 3.8567713542708543e-07, "loss": 0.001, "step": 24046 }, { "epoch": 7.46, "learning_rate": 3.8527705541108227e-07, "loss": 0.0006, "step": 24047 }, { "epoch": 7.46, "learning_rate": 3.8487697539507906e-07, "loss": 0.001, "step": 24048 }, { "epoch": 7.46, "learning_rate": 3.844768953790758e-07, "loss": 0.0006, "step": 24049 }, { "epoch": 7.46, "learning_rate": 3.8407681536307265e-07, "loss": 0.0132, "step": 24050 }, { "epoch": 7.46, "learning_rate": 3.8367673534706944e-07, "loss": 0.0008, "step": 24051 }, { "epoch": 7.47, "learning_rate": 3.8327665533106623e-07, "loss": 0.001, "step": 24052 }, { "epoch": 7.47, "learning_rate": 3.82876575315063e-07, "loss": 0.0014, "step": 24053 }, { "epoch": 7.47, "learning_rate": 3.8247649529905987e-07, "loss": 0.0004, "step": 24054 }, { "epoch": 7.47, "learning_rate": 3.8207641528305666e-07, "loss": 0.0011, "step": 24055 }, { "epoch": 7.47, "learning_rate": 3.816763352670534e-07, "loss": 0.0223, "step": 24056 }, { "epoch": 7.47, "learning_rate": 3.8127625525105025e-07, "loss": 0.0014, "step": 24057 }, { "epoch": 7.47, "learning_rate": 3.8087617523504704e-07, "loss": 0.0006, "step": 24058 }, { "epoch": 7.47, "learning_rate": 3.804760952190439e-07, "loss": 0.0074, "step": 24059 }, { "epoch": 7.47, "learning_rate": 3.800760152030406e-07, "loss": 0.0013, "step": 24060 }, { "epoch": 7.47, "learning_rate": 3.796759351870374e-07, "loss": 0.0006, "step": 24061 }, { "epoch": 7.47, "learning_rate": 3.7927585517103426e-07, "loss": 0.0024, "step": 24062 }, { "epoch": 7.47, "learning_rate": 3.78875775155031e-07, "loss": 0.0007, "step": 24063 }, { "epoch": 7.47, "learning_rate": 3.7847569513902784e-07, "loss": 0.0017, "step": 24064 }, { "epoch": 7.47, "learning_rate": 3.7807561512302464e-07, "loss": 0.0009, "step": 24065 }, { "epoch": 7.47, "learning_rate": 3.776755351070215e-07, "loss": 0.0008, "step": 24066 }, { "epoch": 7.47, "learning_rate": 3.772754550910182e-07, "loss": 0.0009, "step": 24067 }, { "epoch": 7.47, "learning_rate": 3.76875375075015e-07, "loss": 0.0009, "step": 24068 }, { "epoch": 7.47, "learning_rate": 3.7647529505901186e-07, "loss": 0.0006, "step": 24069 }, { "epoch": 7.47, "learning_rate": 3.7607521504300865e-07, "loss": 0.0211, "step": 24070 }, { "epoch": 7.47, "learning_rate": 3.756751350270054e-07, "loss": 0.0009, "step": 24071 }, { "epoch": 7.47, "learning_rate": 3.7527505501100223e-07, "loss": 0.0011, "step": 24072 }, { "epoch": 7.47, "learning_rate": 3.7487497499499903e-07, "loss": 0.0008, "step": 24073 }, { "epoch": 7.47, "learning_rate": 3.7447489497899587e-07, "loss": 0.0005, "step": 24074 }, { "epoch": 7.47, "learning_rate": 3.740748149629926e-07, "loss": 0.0022, "step": 24075 }, { "epoch": 7.47, "learning_rate": 3.736747349469894e-07, "loss": 0.0007, "step": 24076 }, { "epoch": 7.47, "learning_rate": 3.7327465493098625e-07, "loss": 0.0006, "step": 24077 }, { "epoch": 7.47, "learning_rate": 3.72874574914983e-07, "loss": 0.0005, "step": 24078 }, { "epoch": 7.47, "learning_rate": 3.7247449489897983e-07, "loss": 0.0009, "step": 24079 }, { "epoch": 7.47, "learning_rate": 3.720744148829766e-07, "loss": 0.0008, "step": 24080 }, { "epoch": 7.47, "learning_rate": 3.7167433486697347e-07, "loss": 0.0007, "step": 24081 }, { "epoch": 7.47, "learning_rate": 3.712742548509702e-07, "loss": 0.0006, "step": 24082 }, { "epoch": 7.47, "learning_rate": 3.70874174834967e-07, "loss": 0.0006, "step": 24083 }, { "epoch": 7.48, "learning_rate": 3.7047409481896385e-07, "loss": 0.0007, "step": 24084 }, { "epoch": 7.48, "learning_rate": 3.7007401480296064e-07, "loss": 0.0021, "step": 24085 }, { "epoch": 7.48, "learning_rate": 3.696739347869574e-07, "loss": 0.0005, "step": 24086 }, { "epoch": 7.48, "learning_rate": 3.692738547709542e-07, "loss": 0.0022, "step": 24087 }, { "epoch": 7.48, "learning_rate": 3.68873774754951e-07, "loss": 0.0005, "step": 24088 }, { "epoch": 7.48, "learning_rate": 3.6847369473894786e-07, "loss": 0.0012, "step": 24089 }, { "epoch": 7.48, "learning_rate": 3.680736147229446e-07, "loss": 0.002, "step": 24090 }, { "epoch": 7.48, "learning_rate": 3.6767353470694144e-07, "loss": 0.0224, "step": 24091 }, { "epoch": 7.48, "learning_rate": 3.6727345469093824e-07, "loss": 0.0006, "step": 24092 }, { "epoch": 7.48, "learning_rate": 3.66873374674935e-07, "loss": 0.0007, "step": 24093 }, { "epoch": 7.48, "learning_rate": 3.664732946589318e-07, "loss": 0.0012, "step": 24094 }, { "epoch": 7.48, "learning_rate": 3.660732146429286e-07, "loss": 0.0008, "step": 24095 }, { "epoch": 7.48, "learning_rate": 3.6567313462692546e-07, "loss": 0.0008, "step": 24096 }, { "epoch": 7.48, "learning_rate": 3.652730546109222e-07, "loss": 0.0009, "step": 24097 }, { "epoch": 7.48, "learning_rate": 3.64872974594919e-07, "loss": 0.0005, "step": 24098 }, { "epoch": 7.48, "learning_rate": 3.6447289457891583e-07, "loss": 0.0006, "step": 24099 }, { "epoch": 7.48, "learning_rate": 3.6407281456291263e-07, "loss": 0.0011, "step": 24100 }, { "epoch": 7.48, "learning_rate": 3.6367273454690937e-07, "loss": 0.0008, "step": 24101 }, { "epoch": 7.48, "learning_rate": 3.632726545309062e-07, "loss": 0.0012, "step": 24102 }, { "epoch": 7.48, "learning_rate": 3.62872574514903e-07, "loss": 0.0009, "step": 24103 }, { "epoch": 7.48, "learning_rate": 3.624724944988998e-07, "loss": 0.0009, "step": 24104 }, { "epoch": 7.48, "learning_rate": 3.620724144828966e-07, "loss": 0.0006, "step": 24105 }, { "epoch": 7.48, "learning_rate": 3.6167233446689343e-07, "loss": 0.0007, "step": 24106 }, { "epoch": 7.48, "learning_rate": 3.612722544508902e-07, "loss": 0.0005, "step": 24107 }, { "epoch": 7.48, "learning_rate": 3.6087217443488696e-07, "loss": 0.0134, "step": 24108 }, { "epoch": 7.48, "learning_rate": 3.604720944188838e-07, "loss": 0.0007, "step": 24109 }, { "epoch": 7.48, "learning_rate": 3.600720144028806e-07, "loss": 0.0015, "step": 24110 }, { "epoch": 7.48, "learning_rate": 3.5967193438687745e-07, "loss": 0.0023, "step": 24111 }, { "epoch": 7.48, "learning_rate": 3.592718543708742e-07, "loss": 0.0018, "step": 24112 }, { "epoch": 7.48, "learning_rate": 3.58871774354871e-07, "loss": 0.0013, "step": 24113 }, { "epoch": 7.48, "learning_rate": 3.584716943388678e-07, "loss": 0.0004, "step": 24114 }, { "epoch": 7.48, "learning_rate": 3.580716143228646e-07, "loss": 0.0007, "step": 24115 }, { "epoch": 7.49, "learning_rate": 3.576715343068614e-07, "loss": 0.0005, "step": 24116 }, { "epoch": 7.49, "learning_rate": 3.572714542908582e-07, "loss": 0.0013, "step": 24117 }, { "epoch": 7.49, "learning_rate": 3.5687137427485504e-07, "loss": 0.0008, "step": 24118 }, { "epoch": 7.49, "learning_rate": 3.564712942588518e-07, "loss": 0.0011, "step": 24119 }, { "epoch": 7.49, "learning_rate": 3.560712142428486e-07, "loss": 0.001, "step": 24120 }, { "epoch": 7.49, "learning_rate": 3.556711342268454e-07, "loss": 0.0218, "step": 24121 }, { "epoch": 7.49, "learning_rate": 3.552710542108422e-07, "loss": 0.0019, "step": 24122 }, { "epoch": 7.49, "learning_rate": 3.5487097419483895e-07, "loss": 0.0012, "step": 24123 }, { "epoch": 7.49, "learning_rate": 3.544708941788358e-07, "loss": 0.0074, "step": 24124 }, { "epoch": 7.49, "learning_rate": 3.540708141628326e-07, "loss": 0.0026, "step": 24125 }, { "epoch": 7.49, "learning_rate": 3.5367073414682943e-07, "loss": 0.0014, "step": 24126 }, { "epoch": 7.49, "learning_rate": 3.532706541308262e-07, "loss": 0.0079, "step": 24127 }, { "epoch": 7.49, "learning_rate": 3.5287057411482297e-07, "loss": 0.0007, "step": 24128 }, { "epoch": 7.49, "learning_rate": 3.524704940988198e-07, "loss": 0.0007, "step": 24129 }, { "epoch": 7.49, "learning_rate": 3.5207041408281655e-07, "loss": 0.0044, "step": 24130 }, { "epoch": 7.49, "learning_rate": 3.516703340668134e-07, "loss": 0.0052, "step": 24131 }, { "epoch": 7.49, "learning_rate": 3.512702540508102e-07, "loss": 0.0012, "step": 24132 }, { "epoch": 7.49, "learning_rate": 3.5087017403480703e-07, "loss": 0.0008, "step": 24133 }, { "epoch": 7.49, "learning_rate": 3.5047009401880377e-07, "loss": 0.0005, "step": 24134 }, { "epoch": 7.49, "learning_rate": 3.5007001400280056e-07, "loss": 0.0004, "step": 24135 }, { "epoch": 7.49, "learning_rate": 3.496699339867974e-07, "loss": 0.0024, "step": 24136 }, { "epoch": 7.49, "learning_rate": 3.492698539707942e-07, "loss": 0.001, "step": 24137 }, { "epoch": 7.49, "learning_rate": 3.4886977395479094e-07, "loss": 0.0189, "step": 24138 }, { "epoch": 7.49, "learning_rate": 3.484696939387878e-07, "loss": 0.0009, "step": 24139 }, { "epoch": 7.49, "learning_rate": 3.480696139227846e-07, "loss": 0.0014, "step": 24140 }, { "epoch": 7.49, "learning_rate": 3.476695339067814e-07, "loss": 0.0007, "step": 24141 }, { "epoch": 7.49, "learning_rate": 3.4726945389077816e-07, "loss": 0.0006, "step": 24142 }, { "epoch": 7.49, "learning_rate": 3.46869373874775e-07, "loss": 0.001, "step": 24143 }, { "epoch": 7.49, "learning_rate": 3.464692938587718e-07, "loss": 0.0005, "step": 24144 }, { "epoch": 7.49, "learning_rate": 3.4606921384276854e-07, "loss": 0.0025, "step": 24145 }, { "epoch": 7.49, "learning_rate": 3.456691338267654e-07, "loss": 0.0014, "step": 24146 }, { "epoch": 7.49, "learning_rate": 3.452690538107622e-07, "loss": 0.0006, "step": 24147 }, { "epoch": 7.5, "learning_rate": 3.44868973794759e-07, "loss": 0.0008, "step": 24148 }, { "epoch": 7.5, "learning_rate": 3.4446889377875576e-07, "loss": 0.001, "step": 24149 }, { "epoch": 7.5, "learning_rate": 3.4406881376275255e-07, "loss": 0.0006, "step": 24150 }, { "epoch": 7.5, "learning_rate": 3.436687337467494e-07, "loss": 0.0004, "step": 24151 }, { "epoch": 7.5, "learning_rate": 3.432686537307462e-07, "loss": 0.0008, "step": 24152 }, { "epoch": 7.5, "learning_rate": 3.42868573714743e-07, "loss": 0.0032, "step": 24153 }, { "epoch": 7.5, "learning_rate": 3.424684936987398e-07, "loss": 0.0005, "step": 24154 }, { "epoch": 7.5, "learning_rate": 3.4206841368273657e-07, "loss": 0.0513, "step": 24155 }, { "epoch": 7.5, "learning_rate": 3.4166833366673336e-07, "loss": 0.0005, "step": 24156 }, { "epoch": 7.5, "learning_rate": 3.4126825365073015e-07, "loss": 0.0012, "step": 24157 }, { "epoch": 7.5, "learning_rate": 3.40868173634727e-07, "loss": 0.001, "step": 24158 }, { "epoch": 7.5, "learning_rate": 3.404680936187238e-07, "loss": 0.0014, "step": 24159 }, { "epoch": 7.5, "learning_rate": 3.4006801360272053e-07, "loss": 0.0005, "step": 24160 }, { "epoch": 7.5, "learning_rate": 3.3966793358671737e-07, "loss": 0.0004, "step": 24161 }, { "epoch": 7.5, "learning_rate": 3.3926785357071416e-07, "loss": 0.0007, "step": 24162 }, { "epoch": 7.5, "learning_rate": 3.38867773554711e-07, "loss": 0.011, "step": 24163 }, { "epoch": 7.5, "learning_rate": 3.3846769353870775e-07, "loss": 0.0021, "step": 24164 }, { "epoch": 7.5, "learning_rate": 3.3806761352270454e-07, "loss": 0.0007, "step": 24165 }, { "epoch": 7.5, "learning_rate": 3.376675335067014e-07, "loss": 0.0004, "step": 24166 }, { "epoch": 7.5, "learning_rate": 3.372674534906982e-07, "loss": 0.0009, "step": 24167 }, { "epoch": 7.5, "learning_rate": 3.3686737347469497e-07, "loss": 0.0016, "step": 24168 }, { "epoch": 7.5, "learning_rate": 3.3646729345869176e-07, "loss": 0.0007, "step": 24169 }, { "epoch": 7.5, "learning_rate": 3.360672134426886e-07, "loss": 0.0007, "step": 24170 }, { "epoch": 7.5, "learning_rate": 3.3566713342668535e-07, "loss": 0.0011, "step": 24171 }, { "epoch": 7.5, "learning_rate": 3.3526705341068214e-07, "loss": 0.0013, "step": 24172 }, { "epoch": 7.5, "learning_rate": 3.34866973394679e-07, "loss": 0.0012, "step": 24173 }, { "epoch": 7.5, "learning_rate": 3.344668933786758e-07, "loss": 0.0008, "step": 24174 }, { "epoch": 7.5, "learning_rate": 3.340668133626725e-07, "loss": 0.0023, "step": 24175 }, { "epoch": 7.5, "learning_rate": 3.3366673334666936e-07, "loss": 0.0004, "step": 24176 }, { "epoch": 7.5, "learning_rate": 3.3326665333066615e-07, "loss": 0.001, "step": 24177 }, { "epoch": 7.5, "learning_rate": 3.32866573314663e-07, "loss": 0.0009, "step": 24178 }, { "epoch": 7.5, "learning_rate": 3.3246649329865974e-07, "loss": 0.0009, "step": 24179 }, { "epoch": 7.5, "learning_rate": 3.3206641328265653e-07, "loss": 0.0016, "step": 24180 }, { "epoch": 7.51, "learning_rate": 3.316663332666534e-07, "loss": 0.0006, "step": 24181 }, { "epoch": 7.51, "learning_rate": 3.3126625325065017e-07, "loss": 0.0011, "step": 24182 }, { "epoch": 7.51, "learning_rate": 3.3086617323464696e-07, "loss": 0.0005, "step": 24183 }, { "epoch": 7.51, "learning_rate": 3.3046609321864375e-07, "loss": 0.0009, "step": 24184 }, { "epoch": 7.51, "learning_rate": 3.300660132026406e-07, "loss": 0.0016, "step": 24185 }, { "epoch": 7.51, "learning_rate": 3.2966593318663733e-07, "loss": 0.0004, "step": 24186 }, { "epoch": 7.51, "learning_rate": 3.2926585317063413e-07, "loss": 0.0011, "step": 24187 }, { "epoch": 7.51, "learning_rate": 3.2886577315463097e-07, "loss": 0.0005, "step": 24188 }, { "epoch": 7.51, "learning_rate": 3.2846569313862776e-07, "loss": 0.0007, "step": 24189 }, { "epoch": 7.51, "learning_rate": 3.280656131226245e-07, "loss": 0.0005, "step": 24190 }, { "epoch": 7.51, "learning_rate": 3.2766553310662135e-07, "loss": 0.0066, "step": 24191 }, { "epoch": 7.51, "learning_rate": 3.2726545309061814e-07, "loss": 0.0009, "step": 24192 }, { "epoch": 7.51, "learning_rate": 3.26865373074615e-07, "loss": 0.0012, "step": 24193 }, { "epoch": 7.51, "learning_rate": 3.264652930586117e-07, "loss": 0.0014, "step": 24194 }, { "epoch": 7.51, "learning_rate": 3.2606521304260857e-07, "loss": 0.0092, "step": 24195 }, { "epoch": 7.51, "learning_rate": 3.2566513302660536e-07, "loss": 0.0014, "step": 24196 }, { "epoch": 7.51, "learning_rate": 3.252650530106021e-07, "loss": 0.0009, "step": 24197 }, { "epoch": 7.51, "learning_rate": 3.2486497299459895e-07, "loss": 0.0011, "step": 24198 }, { "epoch": 7.51, "learning_rate": 3.2446489297859574e-07, "loss": 0.0006, "step": 24199 }, { "epoch": 7.51, "learning_rate": 3.240648129625926e-07, "loss": 0.0006, "step": 24200 }, { "epoch": 7.51, "learning_rate": 3.236647329465893e-07, "loss": 0.001, "step": 24201 }, { "epoch": 7.51, "learning_rate": 3.232646529305861e-07, "loss": 0.0005, "step": 24202 }, { "epoch": 7.51, "learning_rate": 3.2286457291458296e-07, "loss": 0.0007, "step": 24203 }, { "epoch": 7.51, "learning_rate": 3.2246449289857975e-07, "loss": 0.001, "step": 24204 }, { "epoch": 7.51, "learning_rate": 3.2206441288257654e-07, "loss": 0.0006, "step": 24205 }, { "epoch": 7.51, "learning_rate": 3.2166433286657334e-07, "loss": 0.0009, "step": 24206 }, { "epoch": 7.51, "learning_rate": 3.2126425285057013e-07, "loss": 0.0014, "step": 24207 }, { "epoch": 7.51, "learning_rate": 3.20864172834567e-07, "loss": 0.0022, "step": 24208 }, { "epoch": 7.51, "learning_rate": 3.204640928185637e-07, "loss": 0.0004, "step": 24209 }, { "epoch": 7.51, "learning_rate": 3.2006401280256056e-07, "loss": 0.0025, "step": 24210 }, { "epoch": 7.51, "learning_rate": 3.1966393278655735e-07, "loss": 0.0197, "step": 24211 }, { "epoch": 7.51, "learning_rate": 3.192638527705541e-07, "loss": 0.0005, "step": 24212 }, { "epoch": 7.52, "learning_rate": 3.1886377275455093e-07, "loss": 0.0007, "step": 24213 }, { "epoch": 7.52, "learning_rate": 3.1846369273854773e-07, "loss": 0.0009, "step": 24214 }, { "epoch": 7.52, "learning_rate": 3.1806361272254457e-07, "loss": 0.0012, "step": 24215 }, { "epoch": 7.52, "learning_rate": 3.176635327065413e-07, "loss": 0.0014, "step": 24216 }, { "epoch": 7.52, "learning_rate": 3.172634526905381e-07, "loss": 0.0008, "step": 24217 }, { "epoch": 7.52, "learning_rate": 3.1686337267453495e-07, "loss": 0.0004, "step": 24218 }, { "epoch": 7.52, "learning_rate": 3.1646329265853174e-07, "loss": 0.0015, "step": 24219 }, { "epoch": 7.52, "learning_rate": 3.1606321264252853e-07, "loss": 0.0008, "step": 24220 }, { "epoch": 7.52, "learning_rate": 3.156631326265253e-07, "loss": 0.0013, "step": 24221 }, { "epoch": 7.52, "learning_rate": 3.1526305261052217e-07, "loss": 0.0011, "step": 24222 }, { "epoch": 7.52, "learning_rate": 3.148629725945189e-07, "loss": 0.0006, "step": 24223 }, { "epoch": 7.52, "learning_rate": 3.144628925785157e-07, "loss": 0.0012, "step": 24224 }, { "epoch": 7.52, "learning_rate": 3.1406281256251255e-07, "loss": 0.001, "step": 24225 }, { "epoch": 7.52, "learning_rate": 3.1366273254650934e-07, "loss": 0.0011, "step": 24226 }, { "epoch": 7.52, "learning_rate": 3.132626525305061e-07, "loss": 0.0006, "step": 24227 }, { "epoch": 7.52, "learning_rate": 3.128625725145029e-07, "loss": 0.0006, "step": 24228 }, { "epoch": 7.52, "learning_rate": 3.124624924984997e-07, "loss": 0.0006, "step": 24229 }, { "epoch": 7.52, "learning_rate": 3.120624124824965e-07, "loss": 0.0007, "step": 24230 }, { "epoch": 7.52, "learning_rate": 3.1166233246649335e-07, "loss": 0.0231, "step": 24231 }, { "epoch": 7.52, "learning_rate": 3.1126225245049014e-07, "loss": 0.0009, "step": 24232 }, { "epoch": 7.52, "learning_rate": 3.1086217243448694e-07, "loss": 0.0012, "step": 24233 }, { "epoch": 7.52, "learning_rate": 3.1046209241848373e-07, "loss": 0.0007, "step": 24234 }, { "epoch": 7.52, "learning_rate": 3.100620124024805e-07, "loss": 0.0011, "step": 24235 }, { "epoch": 7.52, "learning_rate": 3.096619323864773e-07, "loss": 0.0013, "step": 24236 }, { "epoch": 7.52, "learning_rate": 3.092618523704741e-07, "loss": 0.005, "step": 24237 }, { "epoch": 7.52, "learning_rate": 3.088617723544709e-07, "loss": 0.001, "step": 24238 }, { "epoch": 7.52, "learning_rate": 3.084616923384677e-07, "loss": 0.0009, "step": 24239 }, { "epoch": 7.52, "learning_rate": 3.0806161232246453e-07, "loss": 0.0036, "step": 24240 }, { "epoch": 7.52, "learning_rate": 3.0766153230646133e-07, "loss": 0.0028, "step": 24241 }, { "epoch": 7.52, "learning_rate": 3.072614522904581e-07, "loss": 0.0017, "step": 24242 }, { "epoch": 7.52, "learning_rate": 3.068613722744549e-07, "loss": 0.0007, "step": 24243 }, { "epoch": 7.52, "learning_rate": 3.064612922584517e-07, "loss": 0.0005, "step": 24244 }, { "epoch": 7.53, "learning_rate": 3.060612122424485e-07, "loss": 0.0006, "step": 24245 }, { "epoch": 7.53, "learning_rate": 3.0566113222644534e-07, "loss": 0.0007, "step": 24246 }, { "epoch": 7.53, "learning_rate": 3.0526105221044213e-07, "loss": 0.0007, "step": 24247 }, { "epoch": 7.53, "learning_rate": 3.048609721944389e-07, "loss": 0.0013, "step": 24248 }, { "epoch": 7.53, "learning_rate": 3.044608921784357e-07, "loss": 0.001, "step": 24249 }, { "epoch": 7.53, "learning_rate": 3.040608121624325e-07, "loss": 0.0009, "step": 24250 }, { "epoch": 7.53, "learning_rate": 3.036607321464293e-07, "loss": 0.0008, "step": 24251 }, { "epoch": 7.53, "learning_rate": 3.032606521304261e-07, "loss": 0.0193, "step": 24252 }, { "epoch": 7.53, "learning_rate": 3.0286057211442294e-07, "loss": 0.0034, "step": 24253 }, { "epoch": 7.53, "learning_rate": 3.024604920984197e-07, "loss": 0.0019, "step": 24254 }, { "epoch": 7.53, "learning_rate": 3.020604120824165e-07, "loss": 0.0007, "step": 24255 }, { "epoch": 7.53, "learning_rate": 3.016603320664133e-07, "loss": 0.0011, "step": 24256 }, { "epoch": 7.53, "learning_rate": 3.012602520504101e-07, "loss": 0.0004, "step": 24257 }, { "epoch": 7.53, "learning_rate": 3.008601720344069e-07, "loss": 0.0011, "step": 24258 }, { "epoch": 7.53, "learning_rate": 3.0046009201840374e-07, "loss": 0.001, "step": 24259 }, { "epoch": 7.53, "learning_rate": 3.000600120024005e-07, "loss": 0.0007, "step": 24260 }, { "epoch": 7.53, "learning_rate": 2.9965993198639733e-07, "loss": 0.0005, "step": 24261 }, { "epoch": 7.53, "learning_rate": 2.992598519703941e-07, "loss": 0.0005, "step": 24262 }, { "epoch": 7.53, "learning_rate": 2.9885977195439086e-07, "loss": 0.0013, "step": 24263 }, { "epoch": 7.53, "learning_rate": 2.984596919383877e-07, "loss": 0.0009, "step": 24264 }, { "epoch": 7.53, "learning_rate": 2.980596119223845e-07, "loss": 0.0013, "step": 24265 }, { "epoch": 7.53, "learning_rate": 2.976595319063813e-07, "loss": 0.0009, "step": 24266 }, { "epoch": 7.53, "learning_rate": 2.972594518903781e-07, "loss": 0.0005, "step": 24267 }, { "epoch": 7.53, "learning_rate": 2.9685937187437493e-07, "loss": 0.001, "step": 24268 }, { "epoch": 7.53, "learning_rate": 2.9645929185837167e-07, "loss": 0.0007, "step": 24269 }, { "epoch": 7.53, "learning_rate": 2.960592118423685e-07, "loss": 0.0008, "step": 24270 }, { "epoch": 7.53, "learning_rate": 2.956591318263653e-07, "loss": 0.001, "step": 24271 }, { "epoch": 7.53, "learning_rate": 2.952590518103621e-07, "loss": 0.001, "step": 24272 }, { "epoch": 7.53, "learning_rate": 2.948589717943589e-07, "loss": 0.0006, "step": 24273 }, { "epoch": 7.53, "learning_rate": 2.9445889177835573e-07, "loss": 0.0007, "step": 24274 }, { "epoch": 7.53, "learning_rate": 2.9405881176235247e-07, "loss": 0.0009, "step": 24275 }, { "epoch": 7.53, "learning_rate": 2.936587317463493e-07, "loss": 0.0025, "step": 24276 }, { "epoch": 7.54, "learning_rate": 2.932586517303461e-07, "loss": 0.0006, "step": 24277 }, { "epoch": 7.54, "learning_rate": 2.928585717143429e-07, "loss": 0.001, "step": 24278 }, { "epoch": 7.54, "learning_rate": 2.924584916983397e-07, "loss": 0.0012, "step": 24279 }, { "epoch": 7.54, "learning_rate": 2.920584116823365e-07, "loss": 0.0007, "step": 24280 }, { "epoch": 7.54, "learning_rate": 2.916583316663333e-07, "loss": 0.0014, "step": 24281 }, { "epoch": 7.54, "learning_rate": 2.9125825165033007e-07, "loss": 0.0013, "step": 24282 }, { "epoch": 7.54, "learning_rate": 2.908581716343269e-07, "loss": 0.0033, "step": 24283 }, { "epoch": 7.54, "learning_rate": 2.904580916183237e-07, "loss": 0.0009, "step": 24284 }, { "epoch": 7.54, "learning_rate": 2.900580116023205e-07, "loss": 0.0205, "step": 24285 }, { "epoch": 7.54, "learning_rate": 2.896579315863173e-07, "loss": 0.0011, "step": 24286 }, { "epoch": 7.54, "learning_rate": 2.892578515703141e-07, "loss": 0.0005, "step": 24287 }, { "epoch": 7.54, "learning_rate": 2.888577715543109e-07, "loss": 0.0006, "step": 24288 }, { "epoch": 7.54, "learning_rate": 2.884576915383077e-07, "loss": 0.003, "step": 24289 }, { "epoch": 7.54, "learning_rate": 2.8805761152230446e-07, "loss": 0.0056, "step": 24290 }, { "epoch": 7.54, "learning_rate": 2.8765753150630125e-07, "loss": 0.0006, "step": 24291 }, { "epoch": 7.54, "learning_rate": 2.872574514902981e-07, "loss": 0.0197, "step": 24292 }, { "epoch": 7.54, "learning_rate": 2.868573714742949e-07, "loss": 0.0004, "step": 24293 }, { "epoch": 7.54, "learning_rate": 2.864572914582917e-07, "loss": 0.0006, "step": 24294 }, { "epoch": 7.54, "learning_rate": 2.860572114422885e-07, "loss": 0.0003, "step": 24295 }, { "epoch": 7.54, "learning_rate": 2.8565713142628527e-07, "loss": 0.0005, "step": 24296 }, { "epoch": 7.54, "learning_rate": 2.8525705141028206e-07, "loss": 0.0008, "step": 24297 }, { "epoch": 7.54, "learning_rate": 2.848569713942789e-07, "loss": 0.0008, "step": 24298 }, { "epoch": 7.54, "learning_rate": 2.844568913782757e-07, "loss": 0.0011, "step": 24299 }, { "epoch": 7.54, "learning_rate": 2.840568113622725e-07, "loss": 0.0008, "step": 24300 }, { "epoch": 7.54, "learning_rate": 2.836567313462693e-07, "loss": 0.0004, "step": 24301 }, { "epoch": 7.54, "learning_rate": 2.8325665133026607e-07, "loss": 0.0011, "step": 24302 }, { "epoch": 7.54, "learning_rate": 2.8285657131426286e-07, "loss": 0.0008, "step": 24303 }, { "epoch": 7.54, "learning_rate": 2.8245649129825966e-07, "loss": 0.0005, "step": 24304 }, { "epoch": 7.54, "learning_rate": 2.820564112822565e-07, "loss": 0.0026, "step": 24305 }, { "epoch": 7.54, "learning_rate": 2.8165633126625324e-07, "loss": 0.0004, "step": 24306 }, { "epoch": 7.54, "learning_rate": 2.812562512502501e-07, "loss": 0.0007, "step": 24307 }, { "epoch": 7.54, "learning_rate": 2.808561712342469e-07, "loss": 0.0012, "step": 24308 }, { "epoch": 7.54, "learning_rate": 2.8045609121824367e-07, "loss": 0.0004, "step": 24309 }, { "epoch": 7.55, "learning_rate": 2.8005601120224046e-07, "loss": 0.0058, "step": 24310 }, { "epoch": 7.55, "learning_rate": 2.796559311862373e-07, "loss": 0.0007, "step": 24311 }, { "epoch": 7.55, "learning_rate": 2.7925585117023405e-07, "loss": 0.0014, "step": 24312 }, { "epoch": 7.55, "learning_rate": 2.788557711542309e-07, "loss": 0.0007, "step": 24313 }, { "epoch": 7.55, "learning_rate": 2.784556911382277e-07, "loss": 0.0011, "step": 24314 }, { "epoch": 7.55, "learning_rate": 2.780556111222245e-07, "loss": 0.0018, "step": 24315 }, { "epoch": 7.55, "learning_rate": 2.7765553110622127e-07, "loss": 0.001, "step": 24316 }, { "epoch": 7.55, "learning_rate": 2.7725545109021806e-07, "loss": 0.001, "step": 24317 }, { "epoch": 7.55, "learning_rate": 2.7685537107421485e-07, "loss": 0.0007, "step": 24318 }, { "epoch": 7.55, "learning_rate": 2.7645529105821164e-07, "loss": 0.0007, "step": 24319 }, { "epoch": 7.55, "learning_rate": 2.760552110422085e-07, "loss": 0.0004, "step": 24320 }, { "epoch": 7.55, "learning_rate": 2.7565513102620523e-07, "loss": 0.001, "step": 24321 }, { "epoch": 7.55, "learning_rate": 2.752550510102021e-07, "loss": 0.004, "step": 24322 }, { "epoch": 7.55, "learning_rate": 2.7485497099419887e-07, "loss": 0.0007, "step": 24323 }, { "epoch": 7.55, "learning_rate": 2.7445489097819566e-07, "loss": 0.0014, "step": 24324 }, { "epoch": 7.55, "learning_rate": 2.7405481096219245e-07, "loss": 0.0004, "step": 24325 }, { "epoch": 7.55, "learning_rate": 2.736547309461893e-07, "loss": 0.0007, "step": 24326 }, { "epoch": 7.55, "learning_rate": 2.7325465093018603e-07, "loss": 0.0008, "step": 24327 }, { "epoch": 7.55, "learning_rate": 2.728545709141829e-07, "loss": 0.0005, "step": 24328 }, { "epoch": 7.55, "learning_rate": 2.7245449089817967e-07, "loss": 0.0006, "step": 24329 }, { "epoch": 7.55, "learning_rate": 2.7205441088217646e-07, "loss": 0.0011, "step": 24330 }, { "epoch": 7.55, "learning_rate": 2.7165433086617326e-07, "loss": 0.0007, "step": 24331 }, { "epoch": 7.55, "learning_rate": 2.7125425085017005e-07, "loss": 0.0018, "step": 24332 }, { "epoch": 7.55, "learning_rate": 2.7085417083416684e-07, "loss": 0.0014, "step": 24333 }, { "epoch": 7.55, "learning_rate": 2.7045409081816363e-07, "loss": 0.0005, "step": 24334 }, { "epoch": 7.55, "learning_rate": 2.700540108021605e-07, "loss": 0.0005, "step": 24335 }, { "epoch": 7.55, "learning_rate": 2.6965393078615727e-07, "loss": 0.001, "step": 24336 }, { "epoch": 7.55, "learning_rate": 2.6925385077015406e-07, "loss": 0.0043, "step": 24337 }, { "epoch": 7.55, "learning_rate": 2.6885377075415085e-07, "loss": 0.0012, "step": 24338 }, { "epoch": 7.55, "learning_rate": 2.6845369073814765e-07, "loss": 0.0008, "step": 24339 }, { "epoch": 7.55, "learning_rate": 2.6805361072214444e-07, "loss": 0.0008, "step": 24340 }, { "epoch": 7.55, "learning_rate": 2.676535307061413e-07, "loss": 0.0009, "step": 24341 }, { "epoch": 7.56, "learning_rate": 2.67253450690138e-07, "loss": 0.0046, "step": 24342 }, { "epoch": 7.56, "learning_rate": 2.668533706741348e-07, "loss": 0.0013, "step": 24343 }, { "epoch": 7.56, "learning_rate": 2.6645329065813166e-07, "loss": 0.0009, "step": 24344 }, { "epoch": 7.56, "learning_rate": 2.6605321064212845e-07, "loss": 0.033, "step": 24345 }, { "epoch": 7.56, "learning_rate": 2.6565313062612524e-07, "loss": 0.0007, "step": 24346 }, { "epoch": 7.56, "learning_rate": 2.6525305061012204e-07, "loss": 0.0104, "step": 24347 }, { "epoch": 7.56, "learning_rate": 2.6485297059411883e-07, "loss": 0.0006, "step": 24348 }, { "epoch": 7.56, "learning_rate": 2.644528905781156e-07, "loss": 0.0007, "step": 24349 }, { "epoch": 7.56, "learning_rate": 2.6405281056211247e-07, "loss": 0.0014, "step": 24350 }, { "epoch": 7.56, "learning_rate": 2.6365273054610926e-07, "loss": 0.0013, "step": 24351 }, { "epoch": 7.56, "learning_rate": 2.6325265053010605e-07, "loss": 0.0007, "step": 24352 }, { "epoch": 7.56, "learning_rate": 2.6285257051410284e-07, "loss": 0.0041, "step": 24353 }, { "epoch": 7.56, "learning_rate": 2.6245249049809963e-07, "loss": 0.0008, "step": 24354 }, { "epoch": 7.56, "learning_rate": 2.6205241048209643e-07, "loss": 0.0012, "step": 24355 }, { "epoch": 7.56, "learning_rate": 2.6165233046609327e-07, "loss": 0.0012, "step": 24356 }, { "epoch": 7.56, "learning_rate": 2.6125225045009006e-07, "loss": 0.0124, "step": 24357 }, { "epoch": 7.56, "learning_rate": 2.608521704340868e-07, "loss": 0.0004, "step": 24358 }, { "epoch": 7.56, "learning_rate": 2.6045209041808365e-07, "loss": 0.0007, "step": 24359 }, { "epoch": 7.56, "learning_rate": 2.6005201040208044e-07, "loss": 0.0015, "step": 24360 }, { "epoch": 7.56, "learning_rate": 2.5965193038607723e-07, "loss": 0.0019, "step": 24361 }, { "epoch": 7.56, "learning_rate": 2.59251850370074e-07, "loss": 0.0009, "step": 24362 }, { "epoch": 7.56, "learning_rate": 2.5885177035407087e-07, "loss": 0.0008, "step": 24363 }, { "epoch": 7.56, "learning_rate": 2.584516903380676e-07, "loss": 0.0007, "step": 24364 }, { "epoch": 7.56, "learning_rate": 2.5805161032206445e-07, "loss": 0.0011, "step": 24365 }, { "epoch": 7.56, "learning_rate": 2.5765153030606125e-07, "loss": 0.0009, "step": 24366 }, { "epoch": 7.56, "learning_rate": 2.5725145029005804e-07, "loss": 0.0007, "step": 24367 }, { "epoch": 7.56, "learning_rate": 2.5685137027405483e-07, "loss": 0.0006, "step": 24368 }, { "epoch": 7.56, "learning_rate": 2.564512902580516e-07, "loss": 0.0006, "step": 24369 }, { "epoch": 7.56, "learning_rate": 2.560512102420484e-07, "loss": 0.0079, "step": 24370 }, { "epoch": 7.56, "learning_rate": 2.556511302260452e-07, "loss": 0.0011, "step": 24371 }, { "epoch": 7.56, "learning_rate": 2.5525105021004205e-07, "loss": 0.0011, "step": 24372 }, { "epoch": 7.56, "learning_rate": 2.548509701940388e-07, "loss": 0.0011, "step": 24373 }, { "epoch": 7.57, "learning_rate": 2.5445089017803564e-07, "loss": 0.0006, "step": 24374 }, { "epoch": 7.57, "learning_rate": 2.5405081016203243e-07, "loss": 0.0008, "step": 24375 }, { "epoch": 7.57, "learning_rate": 2.536507301460292e-07, "loss": 0.0007, "step": 24376 }, { "epoch": 7.57, "learning_rate": 2.53250650130026e-07, "loss": 0.0006, "step": 24377 }, { "epoch": 7.57, "learning_rate": 2.5285057011402286e-07, "loss": 0.0011, "step": 24378 }, { "epoch": 7.57, "learning_rate": 2.524504900980196e-07, "loss": 0.0012, "step": 24379 }, { "epoch": 7.57, "learning_rate": 2.5205041008201644e-07, "loss": 0.0006, "step": 24380 }, { "epoch": 7.57, "learning_rate": 2.5165033006601323e-07, "loss": 0.003, "step": 24381 }, { "epoch": 7.57, "learning_rate": 2.5125025005001e-07, "loss": 0.002, "step": 24382 }, { "epoch": 7.57, "learning_rate": 2.508501700340068e-07, "loss": 0.0006, "step": 24383 }, { "epoch": 7.57, "learning_rate": 2.504500900180036e-07, "loss": 0.0008, "step": 24384 }, { "epoch": 7.57, "learning_rate": 2.500500100020004e-07, "loss": 0.0012, "step": 24385 }, { "epoch": 7.57, "learning_rate": 2.496499299859972e-07, "loss": 0.0014, "step": 24386 }, { "epoch": 7.57, "learning_rate": 2.4924984996999404e-07, "loss": 0.0006, "step": 24387 }, { "epoch": 7.57, "learning_rate": 2.4884976995399083e-07, "loss": 0.0072, "step": 24388 }, { "epoch": 7.57, "learning_rate": 2.484496899379876e-07, "loss": 0.0019, "step": 24389 }, { "epoch": 7.57, "learning_rate": 2.480496099219844e-07, "loss": 0.0006, "step": 24390 }, { "epoch": 7.57, "learning_rate": 2.476495299059812e-07, "loss": 0.0012, "step": 24391 }, { "epoch": 7.57, "learning_rate": 2.47249449889978e-07, "loss": 0.0006, "step": 24392 }, { "epoch": 7.57, "learning_rate": 2.4684936987397485e-07, "loss": 0.0093, "step": 24393 }, { "epoch": 7.57, "learning_rate": 2.464492898579716e-07, "loss": 0.0128, "step": 24394 }, { "epoch": 7.57, "learning_rate": 2.4604920984196843e-07, "loss": 0.0013, "step": 24395 }, { "epoch": 7.57, "learning_rate": 2.456491298259652e-07, "loss": 0.0007, "step": 24396 }, { "epoch": 7.57, "learning_rate": 2.45249049809962e-07, "loss": 0.0078, "step": 24397 }, { "epoch": 7.57, "learning_rate": 2.448489697939588e-07, "loss": 0.0041, "step": 24398 }, { "epoch": 7.57, "learning_rate": 2.444488897779556e-07, "loss": 0.0031, "step": 24399 }, { "epoch": 7.57, "learning_rate": 2.440488097619524e-07, "loss": 0.0136, "step": 24400 }, { "epoch": 7.57, "learning_rate": 2.436487297459492e-07, "loss": 0.001, "step": 24401 }, { "epoch": 7.57, "learning_rate": 2.4324864972994603e-07, "loss": 0.0009, "step": 24402 }, { "epoch": 7.57, "learning_rate": 2.428485697139428e-07, "loss": 0.0012, "step": 24403 }, { "epoch": 7.57, "learning_rate": 2.424484896979396e-07, "loss": 0.0007, "step": 24404 }, { "epoch": 7.57, "learning_rate": 2.420484096819364e-07, "loss": 0.0063, "step": 24405 }, { "epoch": 7.58, "learning_rate": 2.416483296659332e-07, "loss": 0.0162, "step": 24406 }, { "epoch": 7.58, "learning_rate": 2.4124824964993e-07, "loss": 0.0007, "step": 24407 }, { "epoch": 7.58, "learning_rate": 2.4084816963392683e-07, "loss": 0.0005, "step": 24408 }, { "epoch": 7.58, "learning_rate": 2.404480896179236e-07, "loss": 0.0086, "step": 24409 }, { "epoch": 7.58, "learning_rate": 2.4004800960192037e-07, "loss": 0.0003, "step": 24410 }, { "epoch": 7.58, "learning_rate": 2.396479295859172e-07, "loss": 0.0198, "step": 24411 }, { "epoch": 7.58, "learning_rate": 2.39247849569914e-07, "loss": 0.0004, "step": 24412 }, { "epoch": 7.58, "learning_rate": 2.388477695539108e-07, "loss": 0.0162, "step": 24413 }, { "epoch": 7.58, "learning_rate": 2.384476895379076e-07, "loss": 0.0009, "step": 24414 }, { "epoch": 7.58, "learning_rate": 2.380476095219044e-07, "loss": 0.0009, "step": 24415 }, { "epoch": 7.58, "learning_rate": 2.376475295059012e-07, "loss": 0.0009, "step": 24416 }, { "epoch": 7.58, "learning_rate": 2.3724744948989802e-07, "loss": 0.0008, "step": 24417 }, { "epoch": 7.58, "learning_rate": 2.3684736947389478e-07, "loss": 0.0011, "step": 24418 }, { "epoch": 7.58, "learning_rate": 2.364472894578916e-07, "loss": 0.0021, "step": 24419 }, { "epoch": 7.58, "learning_rate": 2.360472094418884e-07, "loss": 0.001, "step": 24420 }, { "epoch": 7.58, "learning_rate": 2.356471294258852e-07, "loss": 0.0003, "step": 24421 }, { "epoch": 7.58, "learning_rate": 2.35247049409882e-07, "loss": 0.0009, "step": 24422 }, { "epoch": 7.58, "learning_rate": 2.3484696939387877e-07, "loss": 0.001, "step": 24423 }, { "epoch": 7.58, "learning_rate": 2.344468893778756e-07, "loss": 0.0115, "step": 24424 }, { "epoch": 7.58, "learning_rate": 2.3404680936187238e-07, "loss": 0.0007, "step": 24425 }, { "epoch": 7.58, "learning_rate": 2.336467293458692e-07, "loss": 0.0006, "step": 24426 }, { "epoch": 7.58, "learning_rate": 2.33246649329866e-07, "loss": 0.001, "step": 24427 }, { "epoch": 7.58, "learning_rate": 2.328465693138628e-07, "loss": 0.0005, "step": 24428 }, { "epoch": 7.58, "learning_rate": 2.3244648929785958e-07, "loss": 0.0295, "step": 24429 }, { "epoch": 7.58, "learning_rate": 2.320464092818564e-07, "loss": 0.0022, "step": 24430 }, { "epoch": 7.58, "learning_rate": 2.3164632926585319e-07, "loss": 0.0011, "step": 24431 }, { "epoch": 7.58, "learning_rate": 2.3124624924985e-07, "loss": 0.0006, "step": 24432 }, { "epoch": 7.58, "learning_rate": 2.308461692338468e-07, "loss": 0.0007, "step": 24433 }, { "epoch": 7.58, "learning_rate": 2.304460892178436e-07, "loss": 0.0018, "step": 24434 }, { "epoch": 7.58, "learning_rate": 2.3004600920184038e-07, "loss": 0.0007, "step": 24435 }, { "epoch": 7.58, "learning_rate": 2.296459291858372e-07, "loss": 0.0011, "step": 24436 }, { "epoch": 7.58, "learning_rate": 2.29245849169834e-07, "loss": 0.001, "step": 24437 }, { "epoch": 7.59, "learning_rate": 2.2884576915383076e-07, "loss": 0.0009, "step": 24438 }, { "epoch": 7.59, "learning_rate": 2.2844568913782758e-07, "loss": 0.0009, "step": 24439 }, { "epoch": 7.59, "learning_rate": 2.2804560912182437e-07, "loss": 0.0013, "step": 24440 }, { "epoch": 7.59, "learning_rate": 2.276455291058212e-07, "loss": 0.001, "step": 24441 }, { "epoch": 7.59, "learning_rate": 2.2724544908981798e-07, "loss": 0.0005, "step": 24442 }, { "epoch": 7.59, "learning_rate": 2.268453690738148e-07, "loss": 0.0009, "step": 24443 }, { "epoch": 7.59, "learning_rate": 2.2644528905781156e-07, "loss": 0.0013, "step": 24444 }, { "epoch": 7.59, "learning_rate": 2.2604520904180838e-07, "loss": 0.0009, "step": 24445 }, { "epoch": 7.59, "learning_rate": 2.2564512902580517e-07, "loss": 0.0006, "step": 24446 }, { "epoch": 7.59, "learning_rate": 2.25245049009802e-07, "loss": 0.0006, "step": 24447 }, { "epoch": 7.59, "learning_rate": 2.2484496899379879e-07, "loss": 0.001, "step": 24448 }, { "epoch": 7.59, "learning_rate": 2.244448889777956e-07, "loss": 0.001, "step": 24449 }, { "epoch": 7.59, "learning_rate": 2.2404480896179237e-07, "loss": 0.014, "step": 24450 }, { "epoch": 7.59, "learning_rate": 2.2364472894578916e-07, "loss": 0.0009, "step": 24451 }, { "epoch": 7.59, "learning_rate": 2.2324464892978598e-07, "loss": 0.0009, "step": 24452 }, { "epoch": 7.59, "learning_rate": 2.2284456891378277e-07, "loss": 0.0008, "step": 24453 }, { "epoch": 7.59, "learning_rate": 2.224444888977796e-07, "loss": 0.0007, "step": 24454 }, { "epoch": 7.59, "learning_rate": 2.2204440888177636e-07, "loss": 0.0008, "step": 24455 }, { "epoch": 7.59, "learning_rate": 2.2164432886577318e-07, "loss": 0.001, "step": 24456 }, { "epoch": 7.59, "learning_rate": 2.2124424884976997e-07, "loss": 0.0008, "step": 24457 }, { "epoch": 7.59, "learning_rate": 2.2084416883376679e-07, "loss": 0.0008, "step": 24458 }, { "epoch": 7.59, "learning_rate": 2.2044408881776358e-07, "loss": 0.0008, "step": 24459 }, { "epoch": 7.59, "learning_rate": 2.200440088017604e-07, "loss": 0.0013, "step": 24460 }, { "epoch": 7.59, "learning_rate": 2.1964392878575716e-07, "loss": 0.0202, "step": 24461 }, { "epoch": 7.59, "learning_rate": 2.1924384876975398e-07, "loss": 0.0006, "step": 24462 }, { "epoch": 7.59, "learning_rate": 2.1884376875375077e-07, "loss": 0.0012, "step": 24463 }, { "epoch": 7.59, "learning_rate": 2.1844368873774754e-07, "loss": 0.0008, "step": 24464 }, { "epoch": 7.59, "learning_rate": 2.1804360872174436e-07, "loss": 0.0004, "step": 24465 }, { "epoch": 7.59, "learning_rate": 2.1764352870574115e-07, "loss": 0.0008, "step": 24466 }, { "epoch": 7.59, "learning_rate": 2.1724344868973797e-07, "loss": 0.0011, "step": 24467 }, { "epoch": 7.59, "learning_rate": 2.1684336867373476e-07, "loss": 0.0023, "step": 24468 }, { "epoch": 7.59, "learning_rate": 2.1644328865773158e-07, "loss": 0.0123, "step": 24469 }, { "epoch": 7.59, "learning_rate": 2.1604320864172834e-07, "loss": 0.0004, "step": 24470 }, { "epoch": 7.6, "learning_rate": 2.1564312862572516e-07, "loss": 0.0005, "step": 24471 }, { "epoch": 7.6, "learning_rate": 2.1524304860972196e-07, "loss": 0.001, "step": 24472 }, { "epoch": 7.6, "learning_rate": 2.1484296859371877e-07, "loss": 0.0047, "step": 24473 }, { "epoch": 7.6, "learning_rate": 2.1444288857771557e-07, "loss": 0.0009, "step": 24474 }, { "epoch": 7.6, "learning_rate": 2.1404280856171239e-07, "loss": 0.0011, "step": 24475 }, { "epoch": 7.6, "learning_rate": 2.1364272854570915e-07, "loss": 0.0009, "step": 24476 }, { "epoch": 7.6, "learning_rate": 2.1324264852970594e-07, "loss": 0.0012, "step": 24477 }, { "epoch": 7.6, "learning_rate": 2.1284256851370276e-07, "loss": 0.0007, "step": 24478 }, { "epoch": 7.6, "learning_rate": 2.1244248849769955e-07, "loss": 0.0156, "step": 24479 }, { "epoch": 7.6, "learning_rate": 2.1204240848169637e-07, "loss": 0.0008, "step": 24480 }, { "epoch": 7.6, "learning_rate": 2.1164232846569314e-07, "loss": 0.0005, "step": 24481 }, { "epoch": 7.6, "learning_rate": 2.1124224844968996e-07, "loss": 0.0018, "step": 24482 }, { "epoch": 7.6, "learning_rate": 2.1084216843368675e-07, "loss": 0.0008, "step": 24483 }, { "epoch": 7.6, "learning_rate": 2.1044208841768357e-07, "loss": 0.001, "step": 24484 }, { "epoch": 7.6, "learning_rate": 2.1004200840168036e-07, "loss": 0.0006, "step": 24485 }, { "epoch": 7.6, "learning_rate": 2.0964192838567718e-07, "loss": 0.0006, "step": 24486 }, { "epoch": 7.6, "learning_rate": 2.0924184836967394e-07, "loss": 0.0008, "step": 24487 }, { "epoch": 7.6, "learning_rate": 2.0884176835367076e-07, "loss": 0.0003, "step": 24488 }, { "epoch": 7.6, "learning_rate": 2.0844168833766755e-07, "loss": 0.001, "step": 24489 }, { "epoch": 7.6, "learning_rate": 2.0804160832166432e-07, "loss": 0.0012, "step": 24490 }, { "epoch": 7.6, "learning_rate": 2.0764152830566114e-07, "loss": 0.001, "step": 24491 }, { "epoch": 7.6, "learning_rate": 2.0724144828965793e-07, "loss": 0.0146, "step": 24492 }, { "epoch": 7.6, "learning_rate": 2.0684136827365475e-07, "loss": 0.0008, "step": 24493 }, { "epoch": 7.6, "learning_rate": 2.0644128825765154e-07, "loss": 0.0007, "step": 24494 }, { "epoch": 7.6, "learning_rate": 2.0604120824164836e-07, "loss": 0.001, "step": 24495 }, { "epoch": 7.6, "learning_rate": 2.0564112822564513e-07, "loss": 0.0008, "step": 24496 }, { "epoch": 7.6, "learning_rate": 2.0524104820964194e-07, "loss": 0.0019, "step": 24497 }, { "epoch": 7.6, "learning_rate": 2.0484096819363874e-07, "loss": 0.0055, "step": 24498 }, { "epoch": 7.6, "learning_rate": 2.0444088817763556e-07, "loss": 0.0021, "step": 24499 }, { "epoch": 7.6, "learning_rate": 2.0404080816163235e-07, "loss": 0.0012, "step": 24500 }, { "epoch": 7.6, "learning_rate": 2.0364072814562917e-07, "loss": 0.001, "step": 24501 }, { "epoch": 7.6, "learning_rate": 2.0324064812962593e-07, "loss": 0.0005, "step": 24502 }, { "epoch": 7.61, "learning_rate": 2.0284056811362272e-07, "loss": 0.0004, "step": 24503 }, { "epoch": 7.61, "learning_rate": 2.0244048809761954e-07, "loss": 0.0009, "step": 24504 }, { "epoch": 7.61, "learning_rate": 2.0204040808161634e-07, "loss": 0.0004, "step": 24505 }, { "epoch": 7.61, "learning_rate": 2.0164032806561315e-07, "loss": 0.0008, "step": 24506 }, { "epoch": 7.61, "learning_rate": 2.0124024804960992e-07, "loss": 0.0005, "step": 24507 }, { "epoch": 7.61, "learning_rate": 2.0084016803360674e-07, "loss": 0.0006, "step": 24508 }, { "epoch": 7.61, "learning_rate": 2.0044008801760353e-07, "loss": 0.0005, "step": 24509 }, { "epoch": 7.61, "learning_rate": 2.0004000800160035e-07, "loss": 0.0009, "step": 24510 }, { "epoch": 7.61, "learning_rate": 1.9963992798559714e-07, "loss": 0.0007, "step": 24511 }, { "epoch": 7.61, "learning_rate": 1.9923984796959396e-07, "loss": 0.0011, "step": 24512 }, { "epoch": 7.61, "learning_rate": 1.9883976795359073e-07, "loss": 0.0007, "step": 24513 }, { "epoch": 7.61, "learning_rate": 1.9843968793758754e-07, "loss": 0.0006, "step": 24514 }, { "epoch": 7.61, "learning_rate": 1.9803960792158434e-07, "loss": 0.0014, "step": 24515 }, { "epoch": 7.61, "learning_rate": 1.9763952790558115e-07, "loss": 0.0005, "step": 24516 }, { "epoch": 7.61, "learning_rate": 1.9723944788957792e-07, "loss": 0.019, "step": 24517 }, { "epoch": 7.61, "learning_rate": 1.968393678735747e-07, "loss": 0.0007, "step": 24518 }, { "epoch": 7.61, "learning_rate": 1.9643928785757153e-07, "loss": 0.0164, "step": 24519 }, { "epoch": 7.61, "learning_rate": 1.9603920784156832e-07, "loss": 0.0008, "step": 24520 }, { "epoch": 7.61, "learning_rate": 1.9563912782556514e-07, "loss": 0.0006, "step": 24521 }, { "epoch": 7.61, "learning_rate": 1.952390478095619e-07, "loss": 0.0006, "step": 24522 }, { "epoch": 7.61, "learning_rate": 1.9483896779355873e-07, "loss": 0.001, "step": 24523 }, { "epoch": 7.61, "learning_rate": 1.9443888777755552e-07, "loss": 0.0051, "step": 24524 }, { "epoch": 7.61, "learning_rate": 1.9403880776155234e-07, "loss": 0.0005, "step": 24525 }, { "epoch": 7.61, "learning_rate": 1.9363872774554913e-07, "loss": 0.0006, "step": 24526 }, { "epoch": 7.61, "learning_rate": 1.9323864772954595e-07, "loss": 0.001, "step": 24527 }, { "epoch": 7.61, "learning_rate": 1.9283856771354271e-07, "loss": 0.0006, "step": 24528 }, { "epoch": 7.61, "learning_rate": 1.9243848769753953e-07, "loss": 0.0008, "step": 24529 }, { "epoch": 7.61, "learning_rate": 1.9203840768153632e-07, "loss": 0.0004, "step": 24530 }, { "epoch": 7.61, "learning_rate": 1.9163832766553312e-07, "loss": 0.0021, "step": 24531 }, { "epoch": 7.61, "learning_rate": 1.9123824764952994e-07, "loss": 0.001, "step": 24532 }, { "epoch": 7.61, "learning_rate": 1.908381676335267e-07, "loss": 0.0008, "step": 24533 }, { "epoch": 7.61, "learning_rate": 1.9043808761752352e-07, "loss": 0.001, "step": 24534 }, { "epoch": 7.62, "learning_rate": 1.900380076015203e-07, "loss": 0.0012, "step": 24535 }, { "epoch": 7.62, "learning_rate": 1.8963792758551713e-07, "loss": 0.0031, "step": 24536 }, { "epoch": 7.62, "learning_rate": 1.8923784756951392e-07, "loss": 0.0011, "step": 24537 }, { "epoch": 7.62, "learning_rate": 1.8883776755351074e-07, "loss": 0.0025, "step": 24538 }, { "epoch": 7.62, "learning_rate": 1.884376875375075e-07, "loss": 0.0006, "step": 24539 }, { "epoch": 7.62, "learning_rate": 1.8803760752150433e-07, "loss": 0.0007, "step": 24540 }, { "epoch": 7.62, "learning_rate": 1.8763752750550112e-07, "loss": 0.0004, "step": 24541 }, { "epoch": 7.62, "learning_rate": 1.8723744748949794e-07, "loss": 0.001, "step": 24542 }, { "epoch": 7.62, "learning_rate": 1.868373674734947e-07, "loss": 0.0011, "step": 24543 }, { "epoch": 7.62, "learning_rate": 1.864372874574915e-07, "loss": 0.0072, "step": 24544 }, { "epoch": 7.62, "learning_rate": 1.860372074414883e-07, "loss": 0.0006, "step": 24545 }, { "epoch": 7.62, "learning_rate": 1.856371274254851e-07, "loss": 0.0015, "step": 24546 }, { "epoch": 7.62, "learning_rate": 1.8523704740948192e-07, "loss": 0.0005, "step": 24547 }, { "epoch": 7.62, "learning_rate": 1.848369673934787e-07, "loss": 0.0007, "step": 24548 }, { "epoch": 7.62, "learning_rate": 1.844368873774755e-07, "loss": 0.0038, "step": 24549 }, { "epoch": 7.62, "learning_rate": 1.840368073614723e-07, "loss": 0.0008, "step": 24550 }, { "epoch": 7.62, "learning_rate": 1.8363672734546912e-07, "loss": 0.0006, "step": 24551 }, { "epoch": 7.62, "learning_rate": 1.832366473294659e-07, "loss": 0.0007, "step": 24552 }, { "epoch": 7.62, "learning_rate": 1.8283656731346273e-07, "loss": 0.0006, "step": 24553 }, { "epoch": 7.62, "learning_rate": 1.824364872974595e-07, "loss": 0.0007, "step": 24554 }, { "epoch": 7.62, "learning_rate": 1.8203640728145631e-07, "loss": 0.0007, "step": 24555 }, { "epoch": 7.62, "learning_rate": 1.816363272654531e-07, "loss": 0.0146, "step": 24556 }, { "epoch": 7.62, "learning_rate": 1.812362472494499e-07, "loss": 0.0005, "step": 24557 }, { "epoch": 7.62, "learning_rate": 1.8083616723344672e-07, "loss": 0.0012, "step": 24558 }, { "epoch": 7.62, "learning_rate": 1.8043608721744348e-07, "loss": 0.0005, "step": 24559 }, { "epoch": 7.62, "learning_rate": 1.800360072014403e-07, "loss": 0.0006, "step": 24560 }, { "epoch": 7.62, "learning_rate": 1.796359271854371e-07, "loss": 0.004, "step": 24561 }, { "epoch": 7.62, "learning_rate": 1.792358471694339e-07, "loss": 0.0006, "step": 24562 }, { "epoch": 7.62, "learning_rate": 1.788357671534307e-07, "loss": 0.0012, "step": 24563 }, { "epoch": 7.62, "learning_rate": 1.7843568713742752e-07, "loss": 0.0007, "step": 24564 }, { "epoch": 7.62, "learning_rate": 1.780356071214243e-07, "loss": 0.0009, "step": 24565 }, { "epoch": 7.62, "learning_rate": 1.776355271054211e-07, "loss": 0.0007, "step": 24566 }, { "epoch": 7.63, "learning_rate": 1.772354470894179e-07, "loss": 0.001, "step": 24567 }, { "epoch": 7.63, "learning_rate": 1.7683536707341472e-07, "loss": 0.0007, "step": 24568 }, { "epoch": 7.63, "learning_rate": 1.7643528705741148e-07, "loss": 0.0011, "step": 24569 }, { "epoch": 7.63, "learning_rate": 1.7603520704140828e-07, "loss": 0.0004, "step": 24570 }, { "epoch": 7.63, "learning_rate": 1.756351270254051e-07, "loss": 0.0013, "step": 24571 }, { "epoch": 7.63, "learning_rate": 1.7523504700940189e-07, "loss": 0.0005, "step": 24572 }, { "epoch": 7.63, "learning_rate": 1.748349669933987e-07, "loss": 0.0013, "step": 24573 }, { "epoch": 7.63, "learning_rate": 1.7443488697739547e-07, "loss": 0.0311, "step": 24574 }, { "epoch": 7.63, "learning_rate": 1.740348069613923e-07, "loss": 0.0007, "step": 24575 }, { "epoch": 7.63, "learning_rate": 1.7363472694538908e-07, "loss": 0.0008, "step": 24576 }, { "epoch": 7.63, "learning_rate": 1.732346469293859e-07, "loss": 0.0006, "step": 24577 }, { "epoch": 7.63, "learning_rate": 1.728345669133827e-07, "loss": 0.0006, "step": 24578 }, { "epoch": 7.63, "learning_rate": 1.724344868973795e-07, "loss": 0.0008, "step": 24579 }, { "epoch": 7.63, "learning_rate": 1.7203440688137628e-07, "loss": 0.0003, "step": 24580 }, { "epoch": 7.63, "learning_rate": 1.716343268653731e-07, "loss": 0.0005, "step": 24581 }, { "epoch": 7.63, "learning_rate": 1.712342468493699e-07, "loss": 0.0005, "step": 24582 }, { "epoch": 7.63, "learning_rate": 1.7083416683336668e-07, "loss": 0.0012, "step": 24583 }, { "epoch": 7.63, "learning_rate": 1.704340868173635e-07, "loss": 0.0009, "step": 24584 }, { "epoch": 7.63, "learning_rate": 1.7003400680136026e-07, "loss": 0.0011, "step": 24585 }, { "epoch": 7.63, "learning_rate": 1.6963392678535708e-07, "loss": 0.0004, "step": 24586 }, { "epoch": 7.63, "learning_rate": 1.6923384676935387e-07, "loss": 0.0009, "step": 24587 }, { "epoch": 7.63, "learning_rate": 1.688337667533507e-07, "loss": 0.002, "step": 24588 }, { "epoch": 7.63, "learning_rate": 1.6843368673734748e-07, "loss": 0.0005, "step": 24589 }, { "epoch": 7.63, "learning_rate": 1.680336067213443e-07, "loss": 0.0011, "step": 24590 }, { "epoch": 7.63, "learning_rate": 1.6763352670534107e-07, "loss": 0.001, "step": 24591 }, { "epoch": 7.63, "learning_rate": 1.672334466893379e-07, "loss": 0.0011, "step": 24592 }, { "epoch": 7.63, "learning_rate": 1.6683336667333468e-07, "loss": 0.0009, "step": 24593 }, { "epoch": 7.63, "learning_rate": 1.664332866573315e-07, "loss": 0.0008, "step": 24594 }, { "epoch": 7.63, "learning_rate": 1.6603320664132826e-07, "loss": 0.0008, "step": 24595 }, { "epoch": 7.63, "learning_rate": 1.6563312662532508e-07, "loss": 0.0017, "step": 24596 }, { "epoch": 7.63, "learning_rate": 1.6523304660932188e-07, "loss": 0.0007, "step": 24597 }, { "epoch": 7.63, "learning_rate": 1.6483296659331867e-07, "loss": 0.0004, "step": 24598 }, { "epoch": 7.63, "learning_rate": 1.6443288657731549e-07, "loss": 0.0008, "step": 24599 }, { "epoch": 7.64, "learning_rate": 1.6403280656131225e-07, "loss": 0.0007, "step": 24600 }, { "epoch": 7.64, "learning_rate": 1.6363272654530907e-07, "loss": 0.0024, "step": 24601 }, { "epoch": 7.64, "learning_rate": 1.6323264652930586e-07, "loss": 0.0022, "step": 24602 }, { "epoch": 7.64, "learning_rate": 1.6283256651330268e-07, "loss": 0.0007, "step": 24603 }, { "epoch": 7.64, "learning_rate": 1.6243248649729947e-07, "loss": 0.0014, "step": 24604 }, { "epoch": 7.64, "learning_rate": 1.620324064812963e-07, "loss": 0.0012, "step": 24605 }, { "epoch": 7.64, "learning_rate": 1.6163232646529306e-07, "loss": 0.0077, "step": 24606 }, { "epoch": 7.64, "learning_rate": 1.6123224644928988e-07, "loss": 0.0034, "step": 24607 }, { "epoch": 7.64, "learning_rate": 1.6083216643328667e-07, "loss": 0.0006, "step": 24608 }, { "epoch": 7.64, "learning_rate": 1.604320864172835e-07, "loss": 0.0005, "step": 24609 }, { "epoch": 7.64, "learning_rate": 1.6003200640128028e-07, "loss": 0.0014, "step": 24610 }, { "epoch": 7.64, "learning_rate": 1.5963192638527704e-07, "loss": 0.0015, "step": 24611 }, { "epoch": 7.64, "learning_rate": 1.5923184636927386e-07, "loss": 0.0011, "step": 24612 }, { "epoch": 7.64, "learning_rate": 1.5883176635327066e-07, "loss": 0.0026, "step": 24613 }, { "epoch": 7.64, "learning_rate": 1.5843168633726747e-07, "loss": 0.0035, "step": 24614 }, { "epoch": 7.64, "learning_rate": 1.5803160632126427e-07, "loss": 0.001, "step": 24615 }, { "epoch": 7.64, "learning_rate": 1.5763152630526108e-07, "loss": 0.0008, "step": 24616 }, { "epoch": 7.64, "learning_rate": 1.5723144628925785e-07, "loss": 0.0005, "step": 24617 }, { "epoch": 7.64, "learning_rate": 1.5683136627325467e-07, "loss": 0.0023, "step": 24618 }, { "epoch": 7.64, "learning_rate": 1.5643128625725146e-07, "loss": 0.0009, "step": 24619 }, { "epoch": 7.64, "learning_rate": 1.5603120624124825e-07, "loss": 0.0082, "step": 24620 }, { "epoch": 7.64, "learning_rate": 1.5563112622524507e-07, "loss": 0.0229, "step": 24621 }, { "epoch": 7.64, "learning_rate": 1.5523104620924186e-07, "loss": 0.001, "step": 24622 }, { "epoch": 7.64, "learning_rate": 1.5483096619323866e-07, "loss": 0.0006, "step": 24623 }, { "epoch": 7.64, "learning_rate": 1.5443088617723545e-07, "loss": 0.0026, "step": 24624 }, { "epoch": 7.64, "learning_rate": 1.5403080616123227e-07, "loss": 0.0016, "step": 24625 }, { "epoch": 7.64, "learning_rate": 1.5363072614522906e-07, "loss": 0.0008, "step": 24626 }, { "epoch": 7.64, "learning_rate": 1.5323064612922585e-07, "loss": 0.0019, "step": 24627 }, { "epoch": 7.64, "learning_rate": 1.5283056611322267e-07, "loss": 0.0028, "step": 24628 }, { "epoch": 7.64, "learning_rate": 1.5243048609721946e-07, "loss": 0.0011, "step": 24629 }, { "epoch": 7.64, "learning_rate": 1.5203040608121625e-07, "loss": 0.001, "step": 24630 }, { "epoch": 7.64, "learning_rate": 1.5163032606521305e-07, "loss": 0.0006, "step": 24631 }, { "epoch": 7.65, "learning_rate": 1.5123024604920984e-07, "loss": 0.0009, "step": 24632 }, { "epoch": 7.65, "learning_rate": 1.5083016603320666e-07, "loss": 0.0006, "step": 24633 }, { "epoch": 7.65, "learning_rate": 1.5043008601720345e-07, "loss": 0.0014, "step": 24634 }, { "epoch": 7.65, "learning_rate": 1.5003000600120024e-07, "loss": 0.0004, "step": 24635 }, { "epoch": 7.65, "learning_rate": 1.4962992598519706e-07, "loss": 0.0004, "step": 24636 }, { "epoch": 7.65, "learning_rate": 1.4922984596919385e-07, "loss": 0.001, "step": 24637 }, { "epoch": 7.65, "learning_rate": 1.4882976595319064e-07, "loss": 0.0009, "step": 24638 }, { "epoch": 7.65, "learning_rate": 1.4842968593718746e-07, "loss": 0.0026, "step": 24639 }, { "epoch": 7.65, "learning_rate": 1.4802960592118426e-07, "loss": 0.0014, "step": 24640 }, { "epoch": 7.65, "learning_rate": 1.4762952590518105e-07, "loss": 0.0017, "step": 24641 }, { "epoch": 7.65, "learning_rate": 1.4722944588917787e-07, "loss": 0.0005, "step": 24642 }, { "epoch": 7.65, "learning_rate": 1.4682936587317466e-07, "loss": 0.0015, "step": 24643 }, { "epoch": 7.65, "learning_rate": 1.4642928585717145e-07, "loss": 0.001, "step": 24644 }, { "epoch": 7.65, "learning_rate": 1.4602920584116824e-07, "loss": 0.0007, "step": 24645 }, { "epoch": 7.65, "learning_rate": 1.4562912582516503e-07, "loss": 0.0012, "step": 24646 }, { "epoch": 7.65, "learning_rate": 1.4522904580916185e-07, "loss": 0.0286, "step": 24647 }, { "epoch": 7.65, "learning_rate": 1.4482896579315865e-07, "loss": 0.0056, "step": 24648 }, { "epoch": 7.65, "learning_rate": 1.4442888577715544e-07, "loss": 0.0203, "step": 24649 }, { "epoch": 7.65, "learning_rate": 1.4402880576115223e-07, "loss": 0.0013, "step": 24650 }, { "epoch": 7.65, "learning_rate": 1.4362872574514905e-07, "loss": 0.0012, "step": 24651 }, { "epoch": 7.65, "learning_rate": 1.4322864572914584e-07, "loss": 0.0049, "step": 24652 }, { "epoch": 7.65, "learning_rate": 1.4282856571314263e-07, "loss": 0.0007, "step": 24653 }, { "epoch": 7.65, "learning_rate": 1.4242848569713945e-07, "loss": 0.0005, "step": 24654 }, { "epoch": 7.65, "learning_rate": 1.4202840568113624e-07, "loss": 0.0005, "step": 24655 }, { "epoch": 7.65, "learning_rate": 1.4162832566513304e-07, "loss": 0.0009, "step": 24656 }, { "epoch": 7.65, "learning_rate": 1.4122824564912983e-07, "loss": 0.0007, "step": 24657 }, { "epoch": 7.65, "learning_rate": 1.4082816563312662e-07, "loss": 0.0007, "step": 24658 }, { "epoch": 7.65, "learning_rate": 1.4042808561712344e-07, "loss": 0.0008, "step": 24659 }, { "epoch": 7.65, "learning_rate": 1.4002800560112023e-07, "loss": 0.0015, "step": 24660 }, { "epoch": 7.65, "learning_rate": 1.3962792558511702e-07, "loss": 0.006, "step": 24661 }, { "epoch": 7.65, "learning_rate": 1.3922784556911384e-07, "loss": 0.0008, "step": 24662 }, { "epoch": 7.65, "learning_rate": 1.3882776555311063e-07, "loss": 0.0006, "step": 24663 }, { "epoch": 7.66, "learning_rate": 1.3842768553710743e-07, "loss": 0.0025, "step": 24664 }, { "epoch": 7.66, "learning_rate": 1.3802760552110424e-07, "loss": 0.001, "step": 24665 }, { "epoch": 7.66, "learning_rate": 1.3762752550510104e-07, "loss": 0.0025, "step": 24666 }, { "epoch": 7.66, "learning_rate": 1.3722744548909783e-07, "loss": 0.0008, "step": 24667 }, { "epoch": 7.66, "learning_rate": 1.3682736547309465e-07, "loss": 0.0005, "step": 24668 }, { "epoch": 7.66, "learning_rate": 1.3642728545709144e-07, "loss": 0.0009, "step": 24669 }, { "epoch": 7.66, "learning_rate": 1.3602720544108823e-07, "loss": 0.0008, "step": 24670 }, { "epoch": 7.66, "learning_rate": 1.3562712542508502e-07, "loss": 0.001, "step": 24671 }, { "epoch": 7.66, "learning_rate": 1.3522704540908182e-07, "loss": 0.0005, "step": 24672 }, { "epoch": 7.66, "learning_rate": 1.3482696539307863e-07, "loss": 0.0011, "step": 24673 }, { "epoch": 7.66, "learning_rate": 1.3442688537707543e-07, "loss": 0.0028, "step": 24674 }, { "epoch": 7.66, "learning_rate": 1.3402680536107222e-07, "loss": 0.0008, "step": 24675 }, { "epoch": 7.66, "learning_rate": 1.33626725345069e-07, "loss": 0.0009, "step": 24676 }, { "epoch": 7.66, "learning_rate": 1.3322664532906583e-07, "loss": 0.0008, "step": 24677 }, { "epoch": 7.66, "learning_rate": 1.3282656531306262e-07, "loss": 0.0064, "step": 24678 }, { "epoch": 7.66, "learning_rate": 1.3242648529705941e-07, "loss": 0.0007, "step": 24679 }, { "epoch": 7.66, "learning_rate": 1.3202640528105623e-07, "loss": 0.0013, "step": 24680 }, { "epoch": 7.66, "learning_rate": 1.3162632526505303e-07, "loss": 0.0211, "step": 24681 }, { "epoch": 7.66, "learning_rate": 1.3122624524904982e-07, "loss": 0.0291, "step": 24682 }, { "epoch": 7.66, "learning_rate": 1.3082616523304664e-07, "loss": 0.0011, "step": 24683 }, { "epoch": 7.66, "learning_rate": 1.304260852170434e-07, "loss": 0.0008, "step": 24684 }, { "epoch": 7.66, "learning_rate": 1.3002600520104022e-07, "loss": 0.0014, "step": 24685 }, { "epoch": 7.66, "learning_rate": 1.29625925185037e-07, "loss": 0.0053, "step": 24686 }, { "epoch": 7.66, "learning_rate": 1.292258451690338e-07, "loss": 0.0008, "step": 24687 }, { "epoch": 7.66, "learning_rate": 1.2882576515303062e-07, "loss": 0.0004, "step": 24688 }, { "epoch": 7.66, "learning_rate": 1.2842568513702742e-07, "loss": 0.0005, "step": 24689 }, { "epoch": 7.66, "learning_rate": 1.280256051210242e-07, "loss": 0.0007, "step": 24690 }, { "epoch": 7.66, "learning_rate": 1.2762552510502103e-07, "loss": 0.0006, "step": 24691 }, { "epoch": 7.66, "learning_rate": 1.2722544508901782e-07, "loss": 0.0019, "step": 24692 }, { "epoch": 7.66, "learning_rate": 1.268253650730146e-07, "loss": 0.0026, "step": 24693 }, { "epoch": 7.66, "learning_rate": 1.2642528505701143e-07, "loss": 0.0032, "step": 24694 }, { "epoch": 7.66, "learning_rate": 1.2602520504100822e-07, "loss": 0.0005, "step": 24695 }, { "epoch": 7.67, "learning_rate": 1.25625125025005e-07, "loss": 0.0011, "step": 24696 }, { "epoch": 7.67, "learning_rate": 1.252250450090018e-07, "loss": 0.0004, "step": 24697 }, { "epoch": 7.67, "learning_rate": 1.248249649929986e-07, "loss": 0.0006, "step": 24698 }, { "epoch": 7.67, "learning_rate": 1.2442488497699542e-07, "loss": 0.0007, "step": 24699 }, { "epoch": 7.67, "learning_rate": 1.240248049609922e-07, "loss": 0.0009, "step": 24700 }, { "epoch": 7.67, "learning_rate": 1.23624724944989e-07, "loss": 0.0008, "step": 24701 }, { "epoch": 7.67, "learning_rate": 1.232246449289858e-07, "loss": 0.0009, "step": 24702 }, { "epoch": 7.67, "learning_rate": 1.228245649129826e-07, "loss": 0.0009, "step": 24703 }, { "epoch": 7.67, "learning_rate": 1.224244848969794e-07, "loss": 0.0006, "step": 24704 }, { "epoch": 7.67, "learning_rate": 1.220244048809762e-07, "loss": 0.0344, "step": 24705 }, { "epoch": 7.67, "learning_rate": 1.2162432486497301e-07, "loss": 0.001, "step": 24706 }, { "epoch": 7.67, "learning_rate": 1.212242448489698e-07, "loss": 0.0005, "step": 24707 }, { "epoch": 7.67, "learning_rate": 1.208241648329666e-07, "loss": 0.0006, "step": 24708 }, { "epoch": 7.67, "learning_rate": 1.2042408481696342e-07, "loss": 0.0004, "step": 24709 }, { "epoch": 7.67, "learning_rate": 1.2002400480096018e-07, "loss": 0.0003, "step": 24710 }, { "epoch": 7.67, "learning_rate": 1.19623924784957e-07, "loss": 0.0009, "step": 24711 }, { "epoch": 7.67, "learning_rate": 1.192238447689538e-07, "loss": 0.0016, "step": 24712 }, { "epoch": 7.67, "learning_rate": 1.188237647529506e-07, "loss": 0.0007, "step": 24713 }, { "epoch": 7.67, "learning_rate": 1.1842368473694739e-07, "loss": 0.0008, "step": 24714 }, { "epoch": 7.67, "learning_rate": 1.180236047209442e-07, "loss": 0.001, "step": 24715 }, { "epoch": 7.67, "learning_rate": 1.17623524704941e-07, "loss": 0.0005, "step": 24716 }, { "epoch": 7.67, "learning_rate": 1.172234446889378e-07, "loss": 0.0014, "step": 24717 }, { "epoch": 7.67, "learning_rate": 1.168233646729346e-07, "loss": 0.0009, "step": 24718 }, { "epoch": 7.67, "learning_rate": 1.164232846569314e-07, "loss": 0.0007, "step": 24719 }, { "epoch": 7.67, "learning_rate": 1.160232046409282e-07, "loss": 0.003, "step": 24720 }, { "epoch": 7.67, "learning_rate": 1.15623124624925e-07, "loss": 0.0007, "step": 24721 }, { "epoch": 7.67, "learning_rate": 1.152230446089218e-07, "loss": 0.0005, "step": 24722 }, { "epoch": 7.67, "learning_rate": 1.148229645929186e-07, "loss": 0.0077, "step": 24723 }, { "epoch": 7.67, "learning_rate": 1.1442288457691538e-07, "loss": 0.0005, "step": 24724 }, { "epoch": 7.67, "learning_rate": 1.1402280456091218e-07, "loss": 0.0006, "step": 24725 }, { "epoch": 7.67, "learning_rate": 1.1362272454490899e-07, "loss": 0.0006, "step": 24726 }, { "epoch": 7.67, "learning_rate": 1.1322264452890578e-07, "loss": 0.0008, "step": 24727 }, { "epoch": 7.68, "learning_rate": 1.1282256451290259e-07, "loss": 0.0014, "step": 24728 }, { "epoch": 7.68, "learning_rate": 1.1242248449689939e-07, "loss": 0.0008, "step": 24729 }, { "epoch": 7.68, "learning_rate": 1.1202240448089618e-07, "loss": 0.0009, "step": 24730 }, { "epoch": 7.68, "learning_rate": 1.1162232446489299e-07, "loss": 0.0007, "step": 24731 }, { "epoch": 7.68, "learning_rate": 1.112222444488898e-07, "loss": 0.0215, "step": 24732 }, { "epoch": 7.68, "learning_rate": 1.1082216443288659e-07, "loss": 0.0007, "step": 24733 }, { "epoch": 7.68, "learning_rate": 1.1042208441688339e-07, "loss": 0.0008, "step": 24734 }, { "epoch": 7.68, "learning_rate": 1.100220044008802e-07, "loss": 0.0016, "step": 24735 }, { "epoch": 7.68, "learning_rate": 1.0962192438487699e-07, "loss": 0.001, "step": 24736 }, { "epoch": 7.68, "learning_rate": 1.0922184436887377e-07, "loss": 0.0009, "step": 24737 }, { "epoch": 7.68, "learning_rate": 1.0882176435287057e-07, "loss": 0.0044, "step": 24738 }, { "epoch": 7.68, "learning_rate": 1.0842168433686738e-07, "loss": 0.0009, "step": 24739 }, { "epoch": 7.68, "learning_rate": 1.0802160432086417e-07, "loss": 0.001, "step": 24740 }, { "epoch": 7.68, "learning_rate": 1.0762152430486098e-07, "loss": 0.0009, "step": 24741 }, { "epoch": 7.68, "learning_rate": 1.0722144428885778e-07, "loss": 0.0011, "step": 24742 }, { "epoch": 7.68, "learning_rate": 1.0682136427285458e-07, "loss": 0.0011, "step": 24743 }, { "epoch": 7.68, "learning_rate": 1.0642128425685138e-07, "loss": 0.0005, "step": 24744 }, { "epoch": 7.68, "learning_rate": 1.0602120424084819e-07, "loss": 0.001, "step": 24745 }, { "epoch": 7.68, "learning_rate": 1.0562112422484498e-07, "loss": 0.0007, "step": 24746 }, { "epoch": 7.68, "learning_rate": 1.0522104420884178e-07, "loss": 0.001, "step": 24747 }, { "epoch": 7.68, "learning_rate": 1.0482096419283859e-07, "loss": 0.0009, "step": 24748 }, { "epoch": 7.68, "learning_rate": 1.0442088417683538e-07, "loss": 0.0007, "step": 24749 }, { "epoch": 7.68, "learning_rate": 1.0402080416083216e-07, "loss": 0.0007, "step": 24750 }, { "epoch": 7.68, "learning_rate": 1.0362072414482897e-07, "loss": 0.0009, "step": 24751 }, { "epoch": 7.68, "learning_rate": 1.0322064412882577e-07, "loss": 0.0006, "step": 24752 }, { "epoch": 7.68, "learning_rate": 1.0282056411282256e-07, "loss": 0.0007, "step": 24753 }, { "epoch": 7.68, "learning_rate": 1.0242048409681937e-07, "loss": 0.0004, "step": 24754 }, { "epoch": 7.68, "learning_rate": 1.0202040408081617e-07, "loss": 0.0012, "step": 24755 }, { "epoch": 7.68, "learning_rate": 1.0162032406481297e-07, "loss": 0.0012, "step": 24756 }, { "epoch": 7.68, "learning_rate": 1.0122024404880977e-07, "loss": 0.0016, "step": 24757 }, { "epoch": 7.68, "learning_rate": 1.0082016403280658e-07, "loss": 0.0008, "step": 24758 }, { "epoch": 7.68, "learning_rate": 1.0042008401680337e-07, "loss": 0.0005, "step": 24759 }, { "epoch": 7.68, "learning_rate": 1.0002000400080017e-07, "loss": 0.0019, "step": 24760 }, { "epoch": 7.69, "learning_rate": 9.961992398479698e-08, "loss": 0.0007, "step": 24761 }, { "epoch": 7.69, "learning_rate": 9.921984396879377e-08, "loss": 0.0005, "step": 24762 }, { "epoch": 7.69, "learning_rate": 9.881976395279058e-08, "loss": 0.001, "step": 24763 }, { "epoch": 7.69, "learning_rate": 9.841968393678736e-08, "loss": 0.0006, "step": 24764 }, { "epoch": 7.69, "learning_rate": 9.801960392078416e-08, "loss": 0.0028, "step": 24765 }, { "epoch": 7.69, "learning_rate": 9.761952390478095e-08, "loss": 0.0009, "step": 24766 }, { "epoch": 7.69, "learning_rate": 9.721944388877776e-08, "loss": 0.0005, "step": 24767 }, { "epoch": 7.69, "learning_rate": 9.681936387277456e-08, "loss": 0.0008, "step": 24768 }, { "epoch": 7.69, "learning_rate": 9.641928385677136e-08, "loss": 0.0007, "step": 24769 }, { "epoch": 7.69, "learning_rate": 9.601920384076816e-08, "loss": 0.0006, "step": 24770 }, { "epoch": 7.69, "learning_rate": 9.561912382476497e-08, "loss": 0.0006, "step": 24771 }, { "epoch": 7.69, "learning_rate": 9.521904380876176e-08, "loss": 0.001, "step": 24772 }, { "epoch": 7.69, "learning_rate": 9.481896379275857e-08, "loss": 0.0009, "step": 24773 }, { "epoch": 7.69, "learning_rate": 9.441888377675537e-08, "loss": 0.0014, "step": 24774 }, { "epoch": 7.69, "learning_rate": 9.401880376075216e-08, "loss": 0.0043, "step": 24775 }, { "epoch": 7.69, "learning_rate": 9.361872374474897e-08, "loss": 0.0009, "step": 24776 }, { "epoch": 7.69, "learning_rate": 9.321864372874575e-08, "loss": 0.0008, "step": 24777 }, { "epoch": 7.69, "learning_rate": 9.281856371274255e-08, "loss": 0.0009, "step": 24778 }, { "epoch": 7.69, "learning_rate": 9.241848369673934e-08, "loss": 0.0008, "step": 24779 }, { "epoch": 7.69, "learning_rate": 9.201840368073615e-08, "loss": 0.0007, "step": 24780 }, { "epoch": 7.69, "learning_rate": 9.161832366473296e-08, "loss": 0.001, "step": 24781 }, { "epoch": 7.69, "learning_rate": 9.121824364872975e-08, "loss": 0.0011, "step": 24782 }, { "epoch": 7.69, "learning_rate": 9.081816363272655e-08, "loss": 0.0054, "step": 24783 }, { "epoch": 7.69, "learning_rate": 9.041808361672336e-08, "loss": 0.0011, "step": 24784 }, { "epoch": 7.69, "learning_rate": 9.001800360072015e-08, "loss": 0.0009, "step": 24785 }, { "epoch": 7.69, "learning_rate": 8.961792358471696e-08, "loss": 0.0008, "step": 24786 }, { "epoch": 7.69, "learning_rate": 8.921784356871376e-08, "loss": 0.0015, "step": 24787 }, { "epoch": 7.69, "learning_rate": 8.881776355271055e-08, "loss": 0.001, "step": 24788 }, { "epoch": 7.69, "learning_rate": 8.841768353670736e-08, "loss": 0.0063, "step": 24789 }, { "epoch": 7.69, "learning_rate": 8.801760352070414e-08, "loss": 0.0004, "step": 24790 }, { "epoch": 7.69, "learning_rate": 8.761752350470094e-08, "loss": 0.0006, "step": 24791 }, { "epoch": 7.69, "learning_rate": 8.721744348869774e-08, "loss": 0.0007, "step": 24792 }, { "epoch": 7.7, "learning_rate": 8.681736347269454e-08, "loss": 0.0009, "step": 24793 }, { "epoch": 7.7, "learning_rate": 8.641728345669135e-08, "loss": 0.0008, "step": 24794 }, { "epoch": 7.7, "learning_rate": 8.601720344068814e-08, "loss": 0.001, "step": 24795 }, { "epoch": 7.7, "learning_rate": 8.561712342468494e-08, "loss": 0.0009, "step": 24796 }, { "epoch": 7.7, "learning_rate": 8.521704340868175e-08, "loss": 0.0006, "step": 24797 }, { "epoch": 7.7, "learning_rate": 8.481696339267854e-08, "loss": 0.0028, "step": 24798 }, { "epoch": 7.7, "learning_rate": 8.441688337667535e-08, "loss": 0.0011, "step": 24799 }, { "epoch": 7.7, "learning_rate": 8.401680336067215e-08, "loss": 0.0006, "step": 24800 }, { "epoch": 7.7, "learning_rate": 8.361672334466894e-08, "loss": 0.0006, "step": 24801 }, { "epoch": 7.7, "learning_rate": 8.321664332866575e-08, "loss": 0.0008, "step": 24802 }, { "epoch": 7.7, "learning_rate": 8.281656331266254e-08, "loss": 0.0186, "step": 24803 }, { "epoch": 7.7, "learning_rate": 8.241648329665933e-08, "loss": 0.0016, "step": 24804 }, { "epoch": 7.7, "learning_rate": 8.201640328065613e-08, "loss": 0.0015, "step": 24805 }, { "epoch": 7.7, "learning_rate": 8.161632326465293e-08, "loss": 0.0011, "step": 24806 }, { "epoch": 7.7, "learning_rate": 8.121624324864974e-08, "loss": 0.0016, "step": 24807 }, { "epoch": 7.7, "learning_rate": 8.081616323264653e-08, "loss": 0.0103, "step": 24808 }, { "epoch": 7.7, "learning_rate": 8.041608321664333e-08, "loss": 0.0007, "step": 24809 }, { "epoch": 7.7, "learning_rate": 8.001600320064014e-08, "loss": 0.0244, "step": 24810 }, { "epoch": 7.7, "learning_rate": 7.961592318463693e-08, "loss": 0.025, "step": 24811 }, { "epoch": 7.7, "learning_rate": 7.921584316863374e-08, "loss": 0.0016, "step": 24812 }, { "epoch": 7.7, "learning_rate": 7.881576315263054e-08, "loss": 0.0006, "step": 24813 }, { "epoch": 7.7, "learning_rate": 7.841568313662733e-08, "loss": 0.0048, "step": 24814 }, { "epoch": 7.7, "learning_rate": 7.801560312062413e-08, "loss": 0.0005, "step": 24815 }, { "epoch": 7.7, "learning_rate": 7.761552310462093e-08, "loss": 0.0011, "step": 24816 }, { "epoch": 7.7, "learning_rate": 7.721544308861772e-08, "loss": 0.0005, "step": 24817 }, { "epoch": 7.7, "learning_rate": 7.681536307261453e-08, "loss": 0.001, "step": 24818 }, { "epoch": 7.7, "learning_rate": 7.641528305661134e-08, "loss": 0.0006, "step": 24819 }, { "epoch": 7.7, "learning_rate": 7.601520304060813e-08, "loss": 0.0005, "step": 24820 }, { "epoch": 7.7, "learning_rate": 7.561512302460492e-08, "loss": 0.0009, "step": 24821 }, { "epoch": 7.7, "learning_rate": 7.521504300860172e-08, "loss": 0.0026, "step": 24822 }, { "epoch": 7.7, "learning_rate": 7.481496299259853e-08, "loss": 0.0006, "step": 24823 }, { "epoch": 7.7, "learning_rate": 7.441488297659532e-08, "loss": 0.0005, "step": 24824 }, { "epoch": 7.71, "learning_rate": 7.401480296059213e-08, "loss": 0.0005, "step": 24825 }, { "epoch": 7.71, "learning_rate": 7.361472294458893e-08, "loss": 0.0015, "step": 24826 }, { "epoch": 7.71, "learning_rate": 7.321464292858573e-08, "loss": 0.0011, "step": 24827 }, { "epoch": 7.71, "learning_rate": 7.281456291258252e-08, "loss": 0.0011, "step": 24828 }, { "epoch": 7.71, "learning_rate": 7.241448289657932e-08, "loss": 0.0482, "step": 24829 }, { "epoch": 7.71, "learning_rate": 7.201440288057611e-08, "loss": 0.0007, "step": 24830 }, { "epoch": 7.71, "learning_rate": 7.161432286457292e-08, "loss": 0.0005, "step": 24831 }, { "epoch": 7.71, "learning_rate": 7.121424284856973e-08, "loss": 0.0007, "step": 24832 }, { "epoch": 7.71, "learning_rate": 7.081416283256652e-08, "loss": 0.0016, "step": 24833 }, { "epoch": 7.71, "learning_rate": 7.041408281656331e-08, "loss": 0.0009, "step": 24834 }, { "epoch": 7.71, "learning_rate": 7.001400280056012e-08, "loss": 0.001, "step": 24835 }, { "epoch": 7.71, "learning_rate": 6.961392278455692e-08, "loss": 0.0006, "step": 24836 }, { "epoch": 7.71, "learning_rate": 6.921384276855371e-08, "loss": 0.0011, "step": 24837 }, { "epoch": 7.71, "learning_rate": 6.881376275255052e-08, "loss": 0.0004, "step": 24838 }, { "epoch": 7.71, "learning_rate": 6.841368273654732e-08, "loss": 0.0011, "step": 24839 }, { "epoch": 7.71, "learning_rate": 6.801360272054412e-08, "loss": 0.0005, "step": 24840 }, { "epoch": 7.71, "learning_rate": 6.761352270454091e-08, "loss": 0.0006, "step": 24841 }, { "epoch": 7.71, "learning_rate": 6.721344268853771e-08, "loss": 0.0008, "step": 24842 }, { "epoch": 7.71, "learning_rate": 6.68133626725345e-08, "loss": 0.001, "step": 24843 }, { "epoch": 7.71, "learning_rate": 6.641328265653131e-08, "loss": 0.0007, "step": 24844 }, { "epoch": 7.71, "learning_rate": 6.601320264052812e-08, "loss": 0.0009, "step": 24845 }, { "epoch": 7.71, "learning_rate": 6.561312262452491e-08, "loss": 0.0009, "step": 24846 }, { "epoch": 7.71, "learning_rate": 6.52130426085217e-08, "loss": 0.0004, "step": 24847 }, { "epoch": 7.71, "learning_rate": 6.48129625925185e-08, "loss": 0.0011, "step": 24848 }, { "epoch": 7.71, "learning_rate": 6.441288257651531e-08, "loss": 0.0006, "step": 24849 }, { "epoch": 7.71, "learning_rate": 6.40128025605121e-08, "loss": 0.0007, "step": 24850 }, { "epoch": 7.71, "learning_rate": 6.361272254450891e-08, "loss": 0.0013, "step": 24851 }, { "epoch": 7.71, "learning_rate": 6.321264252850571e-08, "loss": 0.001, "step": 24852 }, { "epoch": 7.71, "learning_rate": 6.28125625125025e-08, "loss": 0.0016, "step": 24853 }, { "epoch": 7.71, "learning_rate": 6.24124824964993e-08, "loss": 0.0005, "step": 24854 }, { "epoch": 7.71, "learning_rate": 6.20124024804961e-08, "loss": 0.0003, "step": 24855 }, { "epoch": 7.71, "learning_rate": 6.16123224644929e-08, "loss": 0.0011, "step": 24856 }, { "epoch": 7.72, "learning_rate": 6.12122424484897e-08, "loss": 0.0026, "step": 24857 }, { "epoch": 7.72, "learning_rate": 6.081216243248651e-08, "loss": 0.0003, "step": 24858 }, { "epoch": 7.72, "learning_rate": 6.04120824164833e-08, "loss": 0.0007, "step": 24859 }, { "epoch": 7.72, "learning_rate": 6.001200240048009e-08, "loss": 0.001, "step": 24860 }, { "epoch": 7.72, "learning_rate": 5.96119223844769e-08, "loss": 0.0008, "step": 24861 }, { "epoch": 7.72, "learning_rate": 5.9211842368473696e-08, "loss": 0.0008, "step": 24862 }, { "epoch": 7.72, "learning_rate": 5.88117623524705e-08, "loss": 0.0008, "step": 24863 }, { "epoch": 7.72, "learning_rate": 5.84116823364673e-08, "loss": 0.0007, "step": 24864 }, { "epoch": 7.72, "learning_rate": 5.80116023204641e-08, "loss": 0.0018, "step": 24865 }, { "epoch": 7.72, "learning_rate": 5.76115223044609e-08, "loss": 0.0042, "step": 24866 }, { "epoch": 7.72, "learning_rate": 5.721144228845769e-08, "loss": 0.0006, "step": 24867 }, { "epoch": 7.72, "learning_rate": 5.6811362272454495e-08, "loss": 0.0005, "step": 24868 }, { "epoch": 7.72, "learning_rate": 5.6411282256451294e-08, "loss": 0.0005, "step": 24869 }, { "epoch": 7.72, "learning_rate": 5.601120224044809e-08, "loss": 0.0007, "step": 24870 }, { "epoch": 7.72, "learning_rate": 5.56111222244449e-08, "loss": 0.0409, "step": 24871 }, { "epoch": 7.72, "learning_rate": 5.5211042208441697e-08, "loss": 0.0011, "step": 24872 }, { "epoch": 7.72, "learning_rate": 5.4810962192438495e-08, "loss": 0.0138, "step": 24873 }, { "epoch": 7.72, "learning_rate": 5.441088217643529e-08, "loss": 0.0006, "step": 24874 }, { "epoch": 7.72, "learning_rate": 5.4010802160432086e-08, "loss": 0.0017, "step": 24875 }, { "epoch": 7.72, "learning_rate": 5.361072214442889e-08, "loss": 0.0008, "step": 24876 }, { "epoch": 7.72, "learning_rate": 5.321064212842569e-08, "loss": 0.0004, "step": 24877 }, { "epoch": 7.72, "learning_rate": 5.281056211242249e-08, "loss": 0.0009, "step": 24878 }, { "epoch": 7.72, "learning_rate": 5.2410482096419295e-08, "loss": 0.0006, "step": 24879 }, { "epoch": 7.72, "learning_rate": 5.201040208041608e-08, "loss": 0.0024, "step": 24880 }, { "epoch": 7.72, "learning_rate": 5.1610322064412885e-08, "loss": 0.0007, "step": 24881 }, { "epoch": 7.72, "learning_rate": 5.1210242048409684e-08, "loss": 0.003, "step": 24882 }, { "epoch": 7.72, "learning_rate": 5.081016203240648e-08, "loss": 0.0005, "step": 24883 }, { "epoch": 7.72, "learning_rate": 5.041008201640329e-08, "loss": 0.0228, "step": 24884 }, { "epoch": 7.72, "learning_rate": 5.001000200040009e-08, "loss": 0.0008, "step": 24885 }, { "epoch": 7.72, "learning_rate": 4.9609921984396886e-08, "loss": 0.0015, "step": 24886 }, { "epoch": 7.72, "learning_rate": 4.920984196839368e-08, "loss": 0.0003, "step": 24887 }, { "epoch": 7.72, "learning_rate": 4.880976195239048e-08, "loss": 0.0006, "step": 24888 }, { "epoch": 7.73, "learning_rate": 4.840968193638728e-08, "loss": 0.0023, "step": 24889 }, { "epoch": 7.73, "learning_rate": 4.800960192038408e-08, "loss": 0.0011, "step": 24890 }, { "epoch": 7.73, "learning_rate": 4.760952190438088e-08, "loss": 0.0005, "step": 24891 }, { "epoch": 7.73, "learning_rate": 4.7209441888377685e-08, "loss": 0.0015, "step": 24892 }, { "epoch": 7.73, "learning_rate": 4.6809361872374484e-08, "loss": 0.0011, "step": 24893 }, { "epoch": 7.73, "learning_rate": 4.6409281856371276e-08, "loss": 0.0032, "step": 24894 }, { "epoch": 7.73, "learning_rate": 4.6009201840368075e-08, "loss": 0.0003, "step": 24895 }, { "epoch": 7.73, "learning_rate": 4.5609121824364874e-08, "loss": 0.0009, "step": 24896 }, { "epoch": 7.73, "learning_rate": 4.520904180836168e-08, "loss": 0.0006, "step": 24897 }, { "epoch": 7.73, "learning_rate": 4.480896179235848e-08, "loss": 0.001, "step": 24898 }, { "epoch": 7.73, "learning_rate": 4.4408881776355277e-08, "loss": 0.0007, "step": 24899 }, { "epoch": 7.73, "learning_rate": 4.400880176035207e-08, "loss": 0.0005, "step": 24900 }, { "epoch": 7.73, "learning_rate": 4.360872174434887e-08, "loss": 0.0013, "step": 24901 }, { "epoch": 7.73, "learning_rate": 4.320864172834567e-08, "loss": 0.0004, "step": 24902 }, { "epoch": 7.73, "learning_rate": 4.280856171234247e-08, "loss": 0.0013, "step": 24903 }, { "epoch": 7.73, "learning_rate": 4.240848169633927e-08, "loss": 0.0008, "step": 24904 }, { "epoch": 7.73, "learning_rate": 4.2008401680336076e-08, "loss": 0.0008, "step": 24905 }, { "epoch": 7.73, "learning_rate": 4.1608321664332875e-08, "loss": 0.0017, "step": 24906 }, { "epoch": 7.73, "learning_rate": 4.120824164832967e-08, "loss": 0.0006, "step": 24907 }, { "epoch": 7.73, "learning_rate": 4.0808161632326466e-08, "loss": 0.001, "step": 24908 }, { "epoch": 7.73, "learning_rate": 4.0408081616323264e-08, "loss": 0.0023, "step": 24909 }, { "epoch": 7.73, "learning_rate": 4.000800160032007e-08, "loss": 0.0006, "step": 24910 }, { "epoch": 7.73, "learning_rate": 3.960792158431687e-08, "loss": 0.0009, "step": 24911 }, { "epoch": 7.73, "learning_rate": 3.920784156831367e-08, "loss": 0.0006, "step": 24912 }, { "epoch": 7.73, "learning_rate": 3.8807761552310466e-08, "loss": 0.0012, "step": 24913 }, { "epoch": 7.73, "learning_rate": 3.8407681536307265e-08, "loss": 0.0012, "step": 24914 }, { "epoch": 7.73, "learning_rate": 3.8007601520304064e-08, "loss": 0.001, "step": 24915 }, { "epoch": 7.73, "learning_rate": 3.760752150430086e-08, "loss": 0.0012, "step": 24916 }, { "epoch": 7.73, "learning_rate": 3.720744148829766e-08, "loss": 0.0007, "step": 24917 }, { "epoch": 7.73, "learning_rate": 3.6807361472294467e-08, "loss": 0.0006, "step": 24918 }, { "epoch": 7.73, "learning_rate": 3.640728145629126e-08, "loss": 0.0008, "step": 24919 }, { "epoch": 7.73, "learning_rate": 3.600720144028806e-08, "loss": 0.0006, "step": 24920 }, { "epoch": 7.73, "learning_rate": 3.560712142428486e-08, "loss": 0.0009, "step": 24921 }, { "epoch": 7.74, "learning_rate": 3.5207041408281655e-08, "loss": 0.0006, "step": 24922 }, { "epoch": 7.74, "learning_rate": 3.480696139227846e-08, "loss": 0.0044, "step": 24923 }, { "epoch": 7.74, "learning_rate": 3.440688137627526e-08, "loss": 0.0004, "step": 24924 }, { "epoch": 7.74, "learning_rate": 3.400680136027206e-08, "loss": 0.0068, "step": 24925 }, { "epoch": 7.74, "learning_rate": 3.360672134426886e-08, "loss": 0.0009, "step": 24926 }, { "epoch": 7.74, "learning_rate": 3.3206641328265656e-08, "loss": 0.0013, "step": 24927 }, { "epoch": 7.74, "learning_rate": 3.2806561312262454e-08, "loss": 0.0007, "step": 24928 }, { "epoch": 7.74, "learning_rate": 3.240648129625925e-08, "loss": 0.0017, "step": 24929 }, { "epoch": 7.74, "learning_rate": 3.200640128025605e-08, "loss": 0.0011, "step": 24930 }, { "epoch": 7.74, "learning_rate": 3.160632126425286e-08, "loss": 0.0254, "step": 24931 }, { "epoch": 7.74, "learning_rate": 3.120624124824965e-08, "loss": 0.0008, "step": 24932 }, { "epoch": 7.74, "learning_rate": 3.080616123224645e-08, "loss": 0.0058, "step": 24933 }, { "epoch": 7.74, "learning_rate": 3.0406081216243254e-08, "loss": 0.0006, "step": 24934 }, { "epoch": 7.74, "learning_rate": 3.0006001200240046e-08, "loss": 0.0213, "step": 24935 }, { "epoch": 7.74, "learning_rate": 2.9605921184236848e-08, "loss": 0.0008, "step": 24936 }, { "epoch": 7.74, "learning_rate": 2.920584116823365e-08, "loss": 0.0046, "step": 24937 }, { "epoch": 7.74, "learning_rate": 2.880576115223045e-08, "loss": 0.0005, "step": 24938 }, { "epoch": 7.74, "learning_rate": 2.8405681136227247e-08, "loss": 0.0012, "step": 24939 }, { "epoch": 7.74, "learning_rate": 2.8005601120224046e-08, "loss": 0.0008, "step": 24940 }, { "epoch": 7.74, "learning_rate": 2.7605521104220848e-08, "loss": 0.0006, "step": 24941 }, { "epoch": 7.74, "learning_rate": 2.7205441088217644e-08, "loss": 0.0011, "step": 24942 }, { "epoch": 7.74, "learning_rate": 2.6805361072214446e-08, "loss": 0.0007, "step": 24943 }, { "epoch": 7.74, "learning_rate": 2.6405281056211245e-08, "loss": 0.0011, "step": 24944 }, { "epoch": 7.74, "learning_rate": 2.600520104020804e-08, "loss": 0.0007, "step": 24945 }, { "epoch": 7.74, "learning_rate": 2.5605121024204842e-08, "loss": 0.0009, "step": 24946 }, { "epoch": 7.74, "learning_rate": 2.5205041008201644e-08, "loss": 0.0244, "step": 24947 }, { "epoch": 7.74, "learning_rate": 2.4804960992198443e-08, "loss": 0.0003, "step": 24948 }, { "epoch": 7.74, "learning_rate": 2.440488097619524e-08, "loss": 0.0008, "step": 24949 }, { "epoch": 7.74, "learning_rate": 2.400480096019204e-08, "loss": 0.0006, "step": 24950 }, { "epoch": 7.74, "learning_rate": 2.3604720944188843e-08, "loss": 0.001, "step": 24951 }, { "epoch": 7.74, "learning_rate": 2.3204640928185638e-08, "loss": 0.0004, "step": 24952 }, { "epoch": 7.74, "learning_rate": 2.2804560912182437e-08, "loss": 0.0012, "step": 24953 }, { "epoch": 7.75, "learning_rate": 2.240448089617924e-08, "loss": 0.0006, "step": 24954 }, { "epoch": 7.75, "learning_rate": 2.2004400880176034e-08, "loss": 0.0006, "step": 24955 }, { "epoch": 7.75, "learning_rate": 2.1604320864172836e-08, "loss": 0.0013, "step": 24956 }, { "epoch": 7.75, "learning_rate": 2.1204240848169635e-08, "loss": 0.0005, "step": 24957 }, { "epoch": 7.75, "learning_rate": 2.0804160832166437e-08, "loss": 0.0014, "step": 24958 }, { "epoch": 7.75, "learning_rate": 2.0404080816163233e-08, "loss": 0.0002, "step": 24959 }, { "epoch": 7.75, "learning_rate": 2.0004000800160035e-08, "loss": 0.0017, "step": 24960 }, { "epoch": 7.75, "learning_rate": 1.9603920784156834e-08, "loss": 0.003, "step": 24961 }, { "epoch": 7.75, "learning_rate": 1.9203840768153632e-08, "loss": 0.0134, "step": 24962 }, { "epoch": 7.75, "learning_rate": 1.880376075215043e-08, "loss": 0.0008, "step": 24963 }, { "epoch": 7.75, "learning_rate": 1.8403680736147233e-08, "loss": 0.0009, "step": 24964 }, { "epoch": 7.75, "learning_rate": 1.800360072014403e-08, "loss": 0.0011, "step": 24965 }, { "epoch": 7.75, "learning_rate": 1.7603520704140828e-08, "loss": 0.0008, "step": 24966 }, { "epoch": 7.75, "learning_rate": 1.720344068813763e-08, "loss": 0.0013, "step": 24967 }, { "epoch": 7.75, "learning_rate": 1.680336067213443e-08, "loss": 0.0006, "step": 24968 }, { "epoch": 7.75, "learning_rate": 1.6403280656131227e-08, "loss": 0.001, "step": 24969 }, { "epoch": 7.75, "learning_rate": 1.6003200640128026e-08, "loss": 0.0008, "step": 24970 }, { "epoch": 7.75, "learning_rate": 1.5603120624124825e-08, "loss": 0.0007, "step": 24971 }, { "epoch": 7.75, "learning_rate": 1.5203040608121627e-08, "loss": 0.0009, "step": 24972 }, { "epoch": 7.75, "learning_rate": 1.4802960592118424e-08, "loss": 0.0151, "step": 24973 }, { "epoch": 7.75, "learning_rate": 1.4402880576115224e-08, "loss": 0.0007, "step": 24974 }, { "epoch": 7.75, "learning_rate": 1.4002800560112023e-08, "loss": 0.0008, "step": 24975 }, { "epoch": 7.75, "learning_rate": 1.3602720544108822e-08, "loss": 0.0007, "step": 24976 }, { "epoch": 7.75, "learning_rate": 1.3202640528105622e-08, "loss": 0.0007, "step": 24977 }, { "epoch": 7.75, "learning_rate": 1.2802560512102421e-08, "loss": 0.0008, "step": 24978 }, { "epoch": 7.75, "learning_rate": 1.2402480496099221e-08, "loss": 0.0013, "step": 24979 }, { "epoch": 7.75, "learning_rate": 1.200240048009602e-08, "loss": 0.0006, "step": 24980 }, { "epoch": 7.75, "learning_rate": 1.1602320464092819e-08, "loss": 0.0021, "step": 24981 }, { "epoch": 7.75, "learning_rate": 1.120224044808962e-08, "loss": 0.0011, "step": 24982 }, { "epoch": 7.75, "learning_rate": 1.0802160432086418e-08, "loss": 0.0009, "step": 24983 }, { "epoch": 7.75, "learning_rate": 1.0402080416083219e-08, "loss": 0.0016, "step": 24984 }, { "epoch": 7.75, "learning_rate": 1.0002000400080017e-08, "loss": 0.0026, "step": 24985 }, { "epoch": 7.76, "learning_rate": 9.601920384076816e-09, "loss": 0.0012, "step": 24986 }, { "epoch": 7.76, "learning_rate": 9.201840368073617e-09, "loss": 0.0008, "step": 24987 }, { "epoch": 7.76, "learning_rate": 8.801760352070414e-09, "loss": 0.0005, "step": 24988 }, { "epoch": 7.76, "learning_rate": 8.401680336067214e-09, "loss": 0.0008, "step": 24989 }, { "epoch": 7.76, "learning_rate": 8.001600320064013e-09, "loss": 0.0006, "step": 24990 }, { "epoch": 7.76, "learning_rate": 7.601520304060813e-09, "loss": 0.0008, "step": 24991 }, { "epoch": 7.76, "learning_rate": 7.201440288057612e-09, "loss": 0.001, "step": 24992 }, { "epoch": 7.76, "learning_rate": 6.801360272054411e-09, "loss": 0.0006, "step": 24993 }, { "epoch": 7.76, "learning_rate": 6.4012802560512105e-09, "loss": 0.0012, "step": 24994 }, { "epoch": 7.76, "learning_rate": 6.00120024004801e-09, "loss": 0.0015, "step": 24995 }, { "epoch": 7.76, "learning_rate": 5.60112022404481e-09, "loss": 0.018, "step": 24996 }, { "epoch": 7.76, "learning_rate": 5.201040208041609e-09, "loss": 0.0006, "step": 24997 }, { "epoch": 7.76, "learning_rate": 4.800960192038408e-09, "loss": 0.0008, "step": 24998 }, { "epoch": 7.76, "learning_rate": 4.400880176035207e-09, "loss": 0.0009, "step": 24999 }, { "epoch": 7.76, "learning_rate": 4.0008001600320065e-09, "loss": 0.0008, "step": 25000 }, { "epoch": 7.76, "eval_loss": 0.180451899766922, "eval_runtime": 7295.4093, "eval_samples_per_second": 1.498, "eval_steps_per_second": 1.498, "eval_wer": 67.7750577995376, "step": 25000 } ], "max_steps": 25000, "num_train_epochs": 8, "total_flos": 1.15434160128e+20, "trial_name": null, "trial_params": null }